!

Counterfactuals-
and-Causal-Explanation-
in-Historiography-
-
-
-
-
-
-
-
-
-
Päivi-Seppälä-(013226554)-
Pro-gradu,-4.9.2012-
University-of-Helsinki-
Department-of-Philosophy,-History,-Culture-and-Art-Studies
Contents!

1!! INTRODUCTION! 1!
2!! THE!DEBATE!ON!WHAT6MIGHT6HAVE6BEENS! 3!

2.1$Counterfactuals$and$counterfactual$reasoning$ 4!
2.2$What$makes$counterfactuals$dubious?$ 5!
2.2.1!Metaphysical!arguments! 5!
2.2.2!Semantic!arguments! 9!
2.2.3!Epistemic!arguments! 10!
2.2.4!Arguments!based!on!disciplinary!identity! 14!

2.3$Why$are$counterfactuals$indispensable$for$a$historian?$ 15!
2.3.1!Eight!reasons!for!using!counterfactuals! 16!
2.3.2!Mysterious!link!between!causal!explanatory!claims!and!counterfactual!claims! 17!
2.3.3!Causation,!causal!relations!and!causal!explanation! 19!
2.3.4!Argument!for!the!use!of!counterfactuals!in!historiographical!causal!explanations! 20!

3! !CONTRASTIVE!COUNTERFACTUAL!THEORY!OF!CAUSAL!EXPLANATION! 22!

3.1$Two$paths$to$the$contrastive$counterfactual$theory$ 23!
3.1.1!Counterfactual!theories!of!causal!explanation! 23!
3.1.2!Contrastive!approach! 24!

3.2$Main$ideas$of$contrastive$counterfactual$account$ 25!
3.2.1!Contrastive!counterfactual!account!of!singular!causal!explanation! 27!
3.2.2!From!causal!claims!to!counterfactuals! 28!
3.2.3!Tracking!invariant!dependency!relations!with!interventions! 29!

4!! CONTRASTIVE!COUNTERFACTUAL!THEORY!IN!HISTORIOGRAPHY! 33!

4.1$Contrastive$explananda(and$contrastive$causal$claims$in$historiography$ 34!
4.1.1!The!contrastive!explanation!of!facts! 34!
4.1.2!Explanatory!selection!in!fixing!the!causal!background! 36!
4.1.3!Negative!causes!and!the!principled!epistemological!objection!reconsidered! 39!
4.1.4!Objectivity!and!subjectivity!in!historiography! 41!

4.2.$Facing$the$competition:$nonEcausalism$and$covering$laws$ 45!
4.2.1!Intentional!explanations!and!causal!explanations! 45!
4.2.2!Debunking!the!covering!law!model! 48!
5!!!EPISTEMOLOGY!AND!METHODOLOGY!OF!COUNTERFACTUAL!INFERENCES! 52!

5.1$Epistemology:$Counterfactual$reasoning$and$the$use$of$evidence$ 52!
5.2$Interventionist$methodology$for$counterfactual$inferences$ 60!
5.2.1!Features!of!interventions! 60!
5.2.2!Intervention!does!not!change!the!explanandum!variable!directly:!the!problem!of!selecting!
sensible!causes! 63!
5.2.3!Intervention!should!not!affect!“the!middlemen”!independently!of!changing!A:!the!problems!
of!complexity!and!nonTmodularity! 69!
5.2.4!Intervention!does!not!correlate!with!other!causes!of!the!explanandum!variable:!the!
problem!of!the!unknown!common!causes! 76!
5.2.5!Intervention!acts!a!switch:!couldTinferences!and!wouldTinferences! 77!

6!! USES!OF!INTERVENTIONIST!COUNTERFACTUALS! 85!

6.1$Causal$weighting$ 85!
6.1.1!Selecting!causes!and!fixing!the!causal!background! 86!
6.1.2!Using!various!contrastive!explananda! 87!
6.1.3!Analyses!on!contingency!and!overdetermination! 87!
6.1.4!Analyses!on!sensitivity! 91!

6.2$Explaining$with$narratives$ 94!
6.2.1!Counterfactual!inference!and!the!construction!of!narratives! 95!

7!! CONCLUSIONS! 99!

REFERENCES! 104!

!
1!! Introduction!
!!!!!!!
If there is such a thing as a sense of reality – and no one will doubt that it has its raison
d’etre – then there must also be something that one can call the sense of possibility.

– Robert Musil

In his novel the Man without Qualities, Robert Musil contemplates insightfully the
human ability to consider possibilities and to imagine how things might have happened
or would happen. Since the ability to think in modal terms is a universal feature of
human cognition, it should not come as a surprise that not even academic research is
immune to this way of thinking. Counterfactual questions on past possibilities tend to
tingle the imagination of a professional historian in the same way as they tingle the
imagination of any of us – it can be hard to resist asking: what if this and that had been
different? It has been doubted, however, whether the human ability to think about the
past possibilities should have any methodological or epistemic role in the work of a
rigorous history scholar. Should a self-respecting historian stick to her sense of the
actual past and save the disturbing what if-questions for the occasions of playing
“parlour-games with might-have-beens”, as the distinguished methodologist E.H Carr
(1964: 97) suggests?
This thesis deals with the use of counterfactuals in historiography1 from the
point of view of causal explanation. It studies the various roles that counterfactuals and
counterfactual reasoning play in causal explanation and causal thinking in
historiography. The aim is to defend the use of counterfactuals in the context of causal
explanation and to create methodological guidelines for that use.
The motivation for the topic stems from the heated debate that has surrounded
the use of counterfactuals in historiography. In this debate, the use of counterfactuals
has been defended on the basis of claiming that causal thinking and causal explanation
are inherently linked to counterfactual reasoning. This claim has not, however, been
fully argued for. There exists a need in the philosophy and methodology of
historiography to explicate in what ways causal explanation and counterfactual

1
I will use the term ’historiography’ to refer to the research on the human past. In order to avoid
confusions the term ’history’ is reserved to denote that past, not the activity of studying it (see e.g. Tucker
2011).

1
reasoning are linked together. This is the point where philosophy of science and its
theories of causal explanation step in.
The theory that is used in this thesis to explicate the role of counterfactuals in
causal explanation is called the contrastive counterfactual theory of causal explanation.
Both Petri Ylikoski (2001; 2007) and James Woodward (2003) have introduced
accounts that can be regarded as versions of this theory. In the context of Woodward’s
account it is also common to talk of interventionist or manipulationist theory of
causation and causal explanation. The motivation to apply the contrastive counterfactual
theory to the debate on counterfactuals comes from two sources. First of all, the theory
has become the standard theory of causal explanation in philosophy of science within
the past decade. In spite of this, there has been relatively little discussion of this theory
in the context of historiography. It is, therefore, high time to correct the situation with
an extensive discussion about the implications of the contrastive counterfactual theory
to historiography. Is the theory able to provide some methodological fruits that
historians can pick when seeking for the causes of historical events? James Woodward
(2003: 313–314) has doubted the applicability of his theory to other areas of
historiography than economic history in few of his remarks. Woodward’s doubts give
further motivation to choose the contrastive counterfactual theory as the theory to be
discussed and applied. Is Woodward right in his doubts? Is causal explanation in
historiography beyond the reach of the contrastive counterfactual theory?
The thesis consists of five discussion chapters. The second chapter introduces
the basic concepts of the thesis such as ‘counterfactuals’ and ‘causal explanation’ and
analyzes the debate of counterfactuals that has churned among historians within the past
decades. In the third chapter, the basic ideas of the contrastive counterfactual theory are
jotted down. Chapters four, five and six are dedicated to applying the contrastive
counterfactual theory to the most puzzling questions in philosophy of historiography in
order to see how far the theory can go in solving these issues. Most of the topics
discussed are illustrated with examples taken mainly from political history, social
history, history of international relations, and economic history. However, the
arguments and conclusions should be applicable to all fields of historiography. The final
closing chapter seeks to give an answer to those critics who question the use of
counterfactuals in historiography, and to provide a toolkit for counterfactual explanatory
reasoning in historiography based on the findings of the previous chapters.

2
2!! The!debate!on!what6might6have6beens!!
!!!!!!!
The question in history is never what must, or what might have taken place, but solely
what the evidence obliges us to conclude did take place.

– Michael Oakeshott (1966[1933]: 139)

The purpose of this chapter is to lay the background to the following chapters by
presenting and analyzing the methodological discussions on historiographical
counterfactuals that philosophy of historiography has witnessed during the last century
or so. These discussions show that not all historians are willing to regard counterfactual
questions as serious research questions. Some – I will call them antagonists of
counterfactuals – even want to weed out counterfactual reasoning from
historiographical research as a methodologically flawed enterprise (Oakeshott
1966[1933]: 139; Carr 1964: 96–97; Croce 1966: 557; Fischer 1971: 18). On the other
hand, the protagonists of counterfactuals are quick to point out that even the toughest
opponents of counterfactuals implicitly engage in counterfactual reasoning in their
historiographical studies (Bulhof 1999; Kaye 2010; Tetlock & Parker, 2006: 18). The
dividing line between the antagonists and protagonists seems to be extremely polarized.
Next I will examine, analyze and evaluate the arguments presented by both sides
of the controversy and try to locate where the polarization stems from. The analysis of
the antagonists’ arguments brings up the worries and problems that any account
defending the use of counterfactuals needs to answer. This chapter provides some
preliminary answers, but the reader is asked to be patient, and wait for the summaries of
the concluding chapter. The final answers to the antagonists’ arguments can be given
only after careful scrutiny of a theory that justifies the use of counterfactuals as a
methodologically sound enterprise. This theory is presented in chapters 3, 4, 5 and 6.
The analysis of the protagonists’ arguments reveals that the protagonists also need a
theory – namely, an explicit theory of causal explanation to strengthen their arguments
against the antagonists. Before discussing these issues, however, let’s have a look at the
most central concepts of the thesis.

3
2.1!Counterfactuals!and!counterfactual!reasoning!
Counterfactual reasoning is no new game in the town of historiography. Tacitus
mourned the early death of Germanicus as follows: “Had he been the sole arbiter of
events, had held the powers and title of King, he would have outstripped Alexander in
military fame as far as he surpassed him in gentleness, in self-command and in other
nobel qualities” (quoted in Tetlock & Belkin 1996: 3). There is no lack of examples if
one wants to elaborate what kind of claims historians make when immersing themselves
in counterfactual reasoning. Here are a few more examples:

 If Themistocles had not planned and led the Athenians’ attack, the Greeks would
have lost the battle of the Salamis in 480 BC (Hanson 2006).
 Had there been no railroad connections, it would not have significantly restricted
development of the American economy in the last half of the nineteenth century
(Fogel 1964).
 If Kennedy had shown more resolve prior to the Cuban missile crisis,
Khrushchev would not have deployed the missiles (Lebow & Gross Stein 1996).

These examples reveal the general idea of counterfactual claims: a counterfactual claim
is a subjunctive conditional in which the antecedent is known or supposed for the
purposes of argument to be false, but in which the consequent is supposed to follow (not
logically!) from the antecedent (Chisholm 1946: 289–290).
Counterfactual reasoning is the term I use to refer both to the activity of
formulating and justifying counterfactual claims and to the way of formulating research
questions in a counterfactual form. When one formulates and justifies counterfactual
claims, one makes counterfactual inferences. Historians quite often approach
counterfactuals by entertaining counterfactual questions starting with “what if” or
“what would have happened”. For instance, one can ask: What if Stalin had been ousted
as general party secretary of the communist party? What would have happened to the
development of the Soviet Union in the 30’s? Pure definitions, however, do not tell
what the point of presenting these questions is. Should one or should one not engage in
counterfactual reasoning in the first place?

4
2.2!What!makes!counterfactuals!dubious?!
When judgment is brought to bear upon a fact, the fact is taken as it is and not as
it might otherwise have been… Historical necessity has to be affirmed and
continually reaffirmed in order to exclude from history the “conditional” which
has no rightful place there. What is forbidden is… the anti-historical and
illogical “if”. Such an “if” arbitrarily divides the course of history into necessary
facts and accidental facts… This is a game which all of us in moments of
distraction or idleness indulge in, when we muse on the way our life would have
turned out if we had not met a certain person… When the attempt is made to
play this sort of game on the field of history, where it is thoroughly out of place,
the effect is too wearisome to be maintained. (Croce 1966: 557.)

The above comment by Benedetto Croce nicely reveals some of the general arguments
that the antagonists of counterfactuals have presented. Croce’s style of argumentation,
with its richness and rhetorical power, also represents quite nicely the way of writing of
many of the 20th century antagonists of counterfactuals whose main message has been
the total rejection of counterfactual reasoning. Croce’s main message is blunt and clear:
historians should concentrate on historical facts that Croce labels as “necessary facts”.
Counterfactual thinking, on the other hand, is methodologically disastrous and should
be restricted to the contemplation of the coincidences of one’s personal life. In what
follows, I will group the arguments of the antagonists of counterfactuals into four
categories: 1) Metaphysical arguments, 2) Semantic arguments, 3) Epistemological
arguments 4) Arguments based on disciplinary identity.

2.2.1!Metaphysical!arguments!
Croce’s outburst against counterfactuals reveals one kind of metaphysical commitment
that a historian might have: the idea of historical events as necessary facts that could not
have turned out any other way than they did and the methodological idea that this
metaphysical commitment should reverberate to the actual research one conducts. In
general, there have been two lines of argument in the methodological literature which
can be characterized as metaphysical:

(M1) Counterfactual reasoning leads the historian to consider the obscurities of the
ontology of possible worlds (Hawthorn 1991: 5–6; Elster 1978: ch. 6). Thus, in
order to have a methodologically sound basis for the use of counterfactuals, we
need a philosophically and conceptually valid ontology of possible worlds.

5
(M2) Counterfactual reasoning lures one into thinking about the age-old riddles of
determinism, contingency, necessity, fate or free will (Fischer 1971: 18; Croce
1966: 557; Carr 1964: 96–97).

The long tradition of philosophy of history and historiography has taken it almost as a
given that a writer with the aim of developing sophisticated methodology of
historiography, should take a stand on the issues of metaphysical necessity,
contingency, possible worlds or free will1. At least this is the conclusion one can infer
from the abundance of methodological literature filled with these questions, although
these questions are sometimes taken up only to dismiss them as making no difference to
the actual work of an historian (e.g. Carr 1964 ch. 4 and 5; Hawthorn 1991: 5–18;
Ferguson 2011[1997]).
It is true that counterfactual claims are linked to a vast bunch of metaphysical
problems such as the ontology of possible worlds and the problems of modal concepts
like ‘necessity’ and ‘contingency’2. Thus, the question to be solved is whether these
metaphysical issues rise up in the context of empirical historiographical research: are
counterfactuals used in historiographical research in such a way that the working
historian is compelled to take a stance in these metaphysical issues? As the arguments
of the next few paragraphs will reveal, I am most willing to go along with E.H Carr
(1964: 96) who suggests that a historian could do without the “inevitable” as a
metaphysical doctrine: metaphysical questions should be left to poets and
metaphysicians since these questions have no a direct effect on the actual work of a
historian. What I do not agree on is Carr’s (ibid.) idea that counterfactual historians
have introduced tough and obscure metaphysical themes into historical research that
should be about factual descriptions and causal explanation that has, according to Carr,
nothing to do with counterfactuals. Therefore, it will be argued that counterfactuals are
a way to analyze historical causal claims and historical explanations – and this analysis
need not bear any immediate inherent connection to metaphysical questions of
determinism and contingency, let alone to possible worlds. Let me elaborate on this.

1
This is not surprising, since, historically speaking, philosophy of historiography has not been separate
from philosophy of history that – unlike contemporary philosophy of historiography - encompasses the
metaphysical issues of the nature of time and history and asks whether history is progressive, continuous
or discontinuous etc. (Vašiček 2011).
2
For further discussion on the ontological status of the possible worlds see, for instance, Lewis (1986a),
and Melia 1998. Kripke (1972) is a classic study on the concept and semantics of necessity. For a general
overview of epistemology of modalities, see Vaidya (2011).

6
 First of all, I want to tackle the question of possible worlds. In philosophy of
science, the idea that causal explanation and theory of causation have something to do
with possible worlds stems from David Lewis’s work (1973a; 1986b; 2000). However,
the issue of possible worlds has not been frequently brought up in methodological
discussions of historians3. The reason for this might be that the theme of possible
worlds is relatively new in the philosophy of science. It might also simply be that
historians shun Lewis’s theory due to its extreme richness in philosophical and technical
detail. In any case, the contemporary discussion of possible worlds has some of its roots
in Lewis’ theory of causation and causal explanation and his idea that the truth
conditions of counterfactuals should be stated in terms of similarity conditions between
possible worlds.
Lewis defines causation as an ancestral of counterfactual dependence in the
following way: c causes e if only if there is a chain c, d1 … dn, e such that each member
of the chain (except c) is counterfactually dependent on the event preceding it. Crudely
put, c causes e if and only if i) if c occurs, e occurs ii) if c did not occur, e would not
occur. These counterfactuals are deemed as true sentences, if i) the worlds where c
occurs and e occurs are more similar, and therefore closer to the actual world than the
ones where c occurs and e fails to occur, ii) the worlds where c and e do not occur are
more similar to the actual world than the ones where c does not occur and e occurs.
Lewis’ original theory of causation and causal explanation has been shown to
face several severe problems – it fails to secure the temporal asymmetry of causal
dependence, and it solves neither the problem of transitivity or the problem of pre-
emption (Menzies 2009). When these problems are viewed in the light of the fact that
Lewis’ theory explicitly rests on the idea of possible worlds that carry their own
baggage of ontological problems, one could indeed draw a desperate conclusion:
dealing with counterfactuals leads to complicated conceptual problems, and in the end,
to severe metaphysical and ontological problems concerning the ontological status of
possible worlds. Luckily, as the next chapter will reveal, Lewis’s analysis is not the only
game in the town of counterfactual theories of causal explanation.
The idea of possible worlds does not need, however, lead to big ontological
questions of the existence of ontological worlds, although Lewis himself is also known
for his modal realism. According to Lewis, possible worlds actually exist. As Kripke

3
See, however, Climo & Howells (1976); Elster 1978: ch. 6; Bulhof (1999: 167); Fearon (1991: 176);
Hawthorn (1991: 21–25) Tetlock & Belkin (1996: 18), Nolan (2011: 10).

7
(1972: 44) points out, possible worlds can be thought of as stipulations and invitations
to imagine some things being different from the way they are (or were) in the actual
world. To Kripke, possible worlds are not actual entities or distant countries that we
find looking through a (metaphysical?) telescope. This would seem to be a reasonable
attitude to adopt in any empirical field of study where the causal claims are based on
evidence and theoretical considerations concerning this world. This does not mean that
one could not use the idea of comparing the similarities between possible worlds as a
conceptual tool when one tries to formulate and assess one’s causal claims. One could,
for instance, say that based on the evidence, it seems that the world where Franz
Ferdinand gets shot on June 28th in Sarajevo and where the First World War breaks out
is closer to the actual world than the one where Franz Ferdinand gets shot and the
World War one does not occur. But it is very important to notice that it is the empirical
evidence that justifies the similarity comparisons between the worlds and in the end the
truth of one’s causal claims. The possible worlds that one speaks about do not do
provide any evidence of their own – they are tools used to assess what kind of claims the
evidence can support.
Even if it is quite easy to avoid the ontological problems related to possible
worlds, one can still ask whether Lewisian possible world semantics is the most useful
tool to approach historians’ counterfactual reasoning. The volume Counterfactual
thought experiments in World Politics points strongly towards the direction that it is not.
The editors Philip E. Tetlock and Aaron Belkin (1996: 18) report in a footnote that they
had asked the contributors of the volume to perform a Stalnaker-Lewis test on the
historical cases they were studying. In other words, the contributors were asked to use
the Lewisian semantics of possible worlds in order to assess the truth of counterfactual
claims they were proposing based on their work, for instance, on the Soviet politics or
U.S policy towards Iran. Tetlock and Belkin, however, report that none of the
contributors was successful in implementing the Stalnaker-Lewis test in their cases. The
empirical evidence and theories that the contributors had, did not simply give them
enough causal knowledge to stipulate possible worlds and to do similarity comparisons
between these worlds and the actual world4.

4
As comparison the other tests that the authors were asked to do included evaluating their counterfactuals
in terms of their clarity, projectability and their consistency with historical facts, theories and statistical
generalizations. The authors found these tests much more useful in assessing their counterfactuals.

8
 Another big metaphysical question that counterfactuals seem to give rise to is
the question of determinism: do things happen the way they do by necessity or is there
real contingency in the world? E.H. Carr (1964: 96–97) is right in pointing out that
historians who explicitly frame their research questions in counterfactual terms seek to
“measure” the contingency of the historical event in question. But, is this
metaphysically so dubious that it should be left to poets, as Carr ends up suggesting?
Not necessarily. For instance, when Robert Fogel (1964) asks the question of the
significance of railroads for the economic growth in the U.S. he is definitely not making
a case for any metaphysical doctrine of the world history and historical progress.
Instead, he is interested in knowing whether railroads were the only route to the
economic prosperity of the U.S. or whether there were other historically possible routes
that might have resulted in the similar level of GDP. In other words, Fogel is trying to
assess how big a role the other factors played in fixing the outcome: if we take railroads
out, would the outcome still be the same? Chapter 6 further deals with the question of
how to approach the issues of contingency and necessity without waking up the
metaphysical boogieman.

2.2.2!Semantic!arguments!!
One way argue against counterfactuals is to claim that counterfactuals lack clear
meaning or that their truth conditions are too vague. James Woodward (2003: 122)
introduces this quite popular argument by bringing up the following slightly humorous
example by Quine:

(SA) If Julius Caesar had been in charge of the U.N. Forces during the Korean War,
then he would have used (a) nuclear weapons or (b) catapults.

It is indeed quite hard to pinpoint on what basis one could decide which one of the
consequents, (a) or (b), is correct. It can also be quite hard to imagine Caesar in the
context of the 20th century world politics. But one can still ask whether this example
works as a good counterexample against counterfactuals that are actually used in
historiographical research. Are the counterfactuals I presented in the introductory
chapter of the thesis on the same level of ridiculousness than Quine’s counterexample to
counterfactual? My answer is negative.

9
 In the last couple of decades, some historians propagating counterfactuals have
tried to establish counterfactual history as a serious methodological movement. In order
to avoid counterexamples like the one by Quine, they have been eager to stress that
counterfactuals should be well-defined (Tetlock & Belkin 1996: 19–21). They insist that
in well-defined counterfactuals the antecedents should be historically plausible in the
light of the evidence, and that there should always be a connection between the
antecedent and consequent that one is justified to infer from the evidence and theoretical
principles. According to this view, Quine’s example is implausible because we cannot
use historiographical evidence to place Ceasar in the context of the 20th century world
politics and to tell how he would have reacted. As it will be shown later, there can be
well-defined counterfactuals in historical research, and therefore, it would be quite
erroneous to regard Quine’s example as a universal dismissal of counterfactuals.
Quine’s example only shows that ridiculous and murky counterfactuals are what they
are: ridiculous and murky. We should, thus, come up with a theory that helps to
distinguish good counterfactuals from bad ones. The following chapters of the thesis,
chapter 5 in particular, will also take up the question to what extent it is possible to
clearly separate good counterfactuals from bad ones.

2.2.3!Epistemic!arguments!
The epistemological argument stems from the following idea: “The field of the historian
is, and must long remain, the discovery and recording of what actually happened”
(Burton Adams 1909: 236). Burton Adams’s comment is based on the presupposition
that historians are epistemically justified to use the evidence available to them to make
inferences that result in interpretations of what happened. This is a presupposition that
most working historians and philosophers of historiography agree on. The question to
be raised, however, is whether it is epistemologically and methodologically justified to
to use evidence to infer things that did not happen. Michael Oakeshott (1966[1933]:
139) does not seem to think so when he states: “The question in history is never what
must, or what might have taken place, but solely what the evidence obliges us to
conclude did take place”.
Oakeshott draws a strict epistemological distinction between the “factual”
inferences and “counterfactual” inferences. Factual inferences are those inferences that
concern the things that happened and the things that can be known. Factual inferences
are epistemically respectable since there is a causal link between the past events and the

10
evidence we now have: the current evidence consists of the traces that have resulted
from the past events and that still remain. The causal link between the past and the
current evidence justifies the use of that evidence for the purposes of inferring what
happened in the past.
According to Oakeshott, counterfactuals, on the other hand, concern alternative
worlds or alternative happenings that did not take place in the past. These are things that
the evidence necessarily stays silent on – there are no direct causal connections between
our evidence and these alternative worlds or the things that did not happen. I believe
that this epistemological distinction between factual and counterfactual motivates those
writers who dismiss counterfactual reasoning as mere flight of fancy, wishful thinking
or at best pure speculation. Counterfactual inferences can simply never have the same
epistemological status than factual claims, since evidence does not support
counterfactual inferences in the same way it supports factual claims.
The principled epistemological objection to counterfactuals can be shortly
formulated as follows:

(PEO) The evidence only “tells” us about the things that did happen. Since
counterfactuals do not concern things that happened – the antecedents and
consequents are by definition things that did not happen – they do provide any
knowledge of past events, and thus do not belong to historiographical research.

As already stated, the objection is based on the assumption that our evidence can only
tell us about the things that the evidence is causally linked to, that is, the things that
have actually happened in the past. This is the basic assumption of all empirical fields
of study. There is simply no causal link to things that did not happen, and therefore our
empirical fields of study do not concern the things that did not happen. Thus, the
question is whether the antecedents and consequents of counterfactuals refer to things
that have actually taken place or do they refer to alternative worlds where we have no
causal and, therefore, no epistemic access. If counterfactuals refer to things that we have
no epistemic access to, I am most willing to agree with the antagonists of
counterfactuals, and claim that counterfactuals on past events are nonsense. If, on the
other hand, our counterfactual claims somehow are linked to things that our evidence
can “speak of”, the protagonists of the counterfactual reasoning have found the way to
escape the principled epistemological objection.

11
 There is, indeed, a way to escape the principled epistemological objection: it is
to argue that counterfactuals are conceptually linked to causal claims that speak of the
actual causal relations of the past. So, if we have evidence on these causal relations, we
have at least some kind of evidence of the related counterfactuals. This line of
argumentation will be discussed together with the principled epistemic objection in
chapter 4.
Even if there was a way to avoid the principled epistemic objection, it would not
clear counterfactuals from all the epistemic doubts, however. Rejecting the principled
epistemic objection means that our evidence can be used to determine, in principle, the
truth of counterfactual claims. The question still remains whether this is possible in
practice. It might be that it is more difficult to use the evidence to support
counterfactual inferences than to answer, for instance, what happened-questions. This
objection is also a valid point, since we have good grounds to think that answering
simple what happened-questions is not always straightforward, either.
First of all, answering what happened-questions requires a lot inferential work,
since historians’ evidence, the traces from the past, does not directly “tell” us what
happened: one can infer what happened only on the basis of reliable evidence. This
means that one needs to make sure that a certain piece of evidence can be reliably used
as evidence for the purpose of making inferences on a past event. Arriving at
conclusions of the reliability of evidence is not straightforward inferential procedure: it
requires getting answers to causal why-questions. In order to assess the reliability of the
evidence, one usually needs to ask: Why is this evidence the way it is? What is the
causal story behind it? Are there pieces of evidence that seem to tell a conflicting story
of the same event, and if so, why? For instance, internal source criticism requires that
when dealing with documentary evidence one asks why this document was created in
the first place. Why did the author of the document write in this way instead of writing
in some other way? Only by answering these questions one can infer what kind of
things a piece of document can be evidence for, whether the evidence should be used to
shed light on the thoughts and values of its author, whether it reveals something
interesting of the social context of its time, or whether the document could be also
treated as a reliable “eye-witness report” of a particular event.
Answering what happened-questions is not complicated only by the complexity
of inferences related to assessing the reliability of evidence, but also the chronic lack of
evidence that historians constantly face. In other words, historians face the problem of

12
underdetermination not only in principle, but also very often in practice. There are cases
where we do not have enough traces from the past that could be used to separate
between competing descriptions of what happened (see Turner 2007 for further
discussion). This might be due to the fact that the relevant evidence has been destroyed
at some point in history or due to the fact that the past event of interest left little or no
traces to begin with. The fact that there is not always enough evidence even to decide
what happened justifies asking whether we can ever have enough evidence to support
counterfactual claims. Based on these considerations one can formulate the modest
epistemic objection as follows:

(MEO) Since the chronic lack of evidence makes reconstructing the past events hard
enough in itself, how could we even dream of having strong evidential support
for counterfactual inferences? Due to the chronic lack of evidence we should
not regard constructing and justifying counterfactuals as the most essential part
of the work of a historian.

Since the antagonists have mainly concentrated on dismissing counterfactual

reasoning as mere flight of imagination and “a parlour-game with might-have-beens”
(Carr 1964: 97), I have not come across this kind of modest epistemological argument
in the comments of antagonists. David Hackett Fischer (1971: 19) is an antagonist, who
comes rather close to presenting the modest epistemological argument when he
criticizes Robert Fogel’s way of using counterfactuals. Nevertheless, Fischer still rejects
the use of counterfactuals for principled reasons: “It is always possible to convert any
historical problem into nonhistorical one, but why should a scholar go out of his way to
make a difficult problem impossible? History is tough enough as it – as it actually is”
(ibid.). While the antagonists have been quite silent on the actual problems that
historians face when using counterfactuals, the protagonists of counterfactuals who
explicitly engage in counterfactual reasoning in their studies have been willing to admit
the problems that the lack of evidence poses for the use of counterfactuals (e.g. see
Lebow & Gross Stein 1996). The modest epistemological objection will be addressed
again in chapter 5 that discusses the use of evidence in supporting counterfactual
claims.

13
2.2.4!Arguments!based!on!disciplinary!identity!
“What is forbidden is… the anti-historical and illogical ‘if’ “(Croce 1966: 557). Croce’s
statement is a representative example of the demarcationist tendency of many
antagonists of counterfactuals. To declare something “anti-historical” or “a-historical”
is a powerful rhetorical tool that is usually used as a knockdown argument to nail the
coffin of counterfactual reasoning for good. Where does this kind of rhetoric originate
from?
Counterfactuals have aroused strong emotions for and against throughout the
whole 20th century (Kaye 2010). This might explain why methodologist have been keen
on making hasty categorical distinctions between “factual” and “counterfactual”
reasoning and also between “history” and “anti-history”. The eagerness to jump into
drastic either-or conclusions has led to the lack of balanced and conciliatory
argumentation between the opposing views. Luckily, after the beginning of the 1990’s
and the publication of Geoffrey Hawthorn’s monograph Plausible Worlds, a slightly
more modest methodological discussion on counterfactuals has emerged.
In the arguments of antagonists, history is contrasted with “anti-historical”
counterfactual reasoning in two ways. Usually the contrast is drawn between the
rigorous history writing that concentrates on the things that actually happened and the
flights of fancy or emotional irrationality that the counterfactual historians are accused
of. In this case, the demarcation between history and “anti-history” is done between an
epistemically respectable systematic field of study and epistemically dubious flights of
fancy that are not based on evidence, but rather on imagination and emotions.
Another demarcation line has been drawn between history as a humanistic study
or a study of man and the natural science (and also social sciences if the social sciences
are considered to imitate the methodologies of natural sciences). Oakeshott (1966
[1933]: 129) states: “ The distinction … between essential and incidental events does
not belong to the history at all: it is a monstrous incursion of science into the world of
history”. Counterfactual reasoning is thus seen as epistemically respectable, but only in
the context of science. If history is seen as distinctively different from the sciences, not
only in respect to the object of study but also methodologically, all the methodological
tools that are believed to be imported from the sciences are treated with doubt or
straightforward rejection. The debate of the 1970’s on Robert Fogel’s work on the
construction of the U.S railroads and the U.S civil war and the abolition of slavery was

14
partly motivated by the strong demarcation between the traditional historiographical
methodology and the (social) scientific methodologies that the cliometric or scientific
movement in economic history was promoting. The movement, and Robert Fogel, in
particular, regarded the use of counterfactuals as central to social scientific causal
reasoning whereas the proponents of the “traditional” doubted the applicability of this
type of reasoning in the context of historiography (Fogel & Elton 1983).
Arguments of disciplinary identity are rhetorically strong, but drastic either-or
distinctions can quite often turn out to be descriptively weak. Many proponents of
counterfactuals have claimed that this observation also applies to the case of
counterfactuals in historiography. They have pointed out that almost all historians, even
the strictest objectors, are relying on counterfactual reasoning in one way or another
(Bulhof 1999; Kaye 2010). According to Tetlock & Parker (2006: 18), the use of verbs
such as influenced, shaped, responded, transformed, triggered, molded, precipitated
takes the historian into the realm of causal claims, and if protagonists are right straight
to counterfactual reasoning. The claim that every historian is, in fact, in one way or
another engaged in counterfactual reasoning is an apt way of pointing out that no one is
actually practicing what the high-fly rhetoric of disciplinary identity is preaching.
However, showing that there is a practice cannot be taken as a justification for why
there should be such a practice. One needs to be able to explain why it is epistemically
and methodologically justifiable and even desirable to engage in counterfactual
reasoning. The following section will deal with this issue of justification.

2.3!Why!are!counterfactuals!indispensable!for!a!historian?!
Throughout the 20th century, counterfactual reasoning has had its supporters (for a
general historical overview, see Kaye 2010). The most common general methodological
argument for counterfactuals has been that counterfactuals are indispensable for
identifying causes, and hence, for explaining events in history (Weber 1949: 164–188;
Aron 1961: 156–186; Fogel 1964; Hawthorn 1991, cp. 2; Fearon 1991, 1996; Tetlock
& Belkin 1996; Stanford 1998: 85–93; Bulhof 1999; Bunzl 2004; Lebow 2007; Megill
2007: 100, 155; Kaye 2010). Despite the prominent role given to the arguments that link
causal explanations to counterfactuals, there are several other arguments presented to
justify the explicit use of counterfactuals in historiography. Thus, before going into a
closer analysis of the arguments on causal explanation, I will briefly introduce Daniel
Nolan’s (2011) list of eight reasons for why historians should use counterfactuals. The

15
list highlights the numerous methodological gains that the explicit use of
counterfactuals brings with it.

2.3.1!Eight!reasons!for!using!counterfactuals!
Daniel Nolan (2011) has provided the most systematic overview of the arguments that
various proponents of counterfactuals, notably Richard Ned Lebow (2007; 2010),
Tetlock and Belkin (1996) and Tetlock & Parker (2006), have presented. Nolan does not
claim that his list is an exhaustive list of all the reasons why counterfactuals are
important in historiography, but in any case Nolan’s suggestions can be said to provide
a good starting point for a more refined research on the matter. Nolan lists in total eight
uses of counterfactuals that he believes to be justified on the basis of the epistemic
concerns of historians and the ways in which historiographical research is conducted.
Four uses are of the sort that one does not need to know the truth of the counterfactuals
when discussing them. Instead, the other four require, according to Nolan, the ability to
assess the truth of the counterfactuals one is dealing with.

Uses of counterfactuals that do not require knowing the truth-values of counterfactuals

1) The use of counterfactuals stimulates historians’ imagination and suggests new

hypotheses to be tested. For instance, when doing a comparative case study on
the feudalisms of the Japanese shogunate and Medieval Latin Europe, it might
be useful to ask how the Japanese shogunate would have dealt with the split of
religious authority. The answer to this question helps to highlight the differences
and similarities of the feudalists systems. (Nolan 2011: 4)
2) Comparing the counterfactuals that the different sides of a historiographical
controversy are willing to accept clarifies what the debating sides are in fact
disagreeing on in their differing causal accounts: the other side might claim that
the absence of this and that factor would have caused the result of interest while
the other side strongly disagrees on this and claims that if the suggested factors
had not been present, the result would still have occurred (Nolan 2011: 5).
3) The explicit use of counterfactuals helps to tackle the so called hindsight bias,
the human tendency to see past events as more inevitable or overdetermined than
they actually were considered to be when they occurred (Nolan 2011: 5; see also
Lebow 2007; Tetlock and Belkin 1996: 15).

16
 4) Describing the thoughts, beliefs and desires of historical actors might sometimes
require that one is interested in the counterfactual reasoning that the actors
themselves engaged in when making decisions (Nolan 2011: 6–7).

Uses of counterfactuals that require assessing their truth-values

5) Thinking in counterfactual terms and trying to assess the truth of counterfactuals

is interesting in its own right (Nolan 2011: 8–9).
6) Counterfactuals are closely connected to causation, and causal reasoning
sometimes seems to require counterfactual considerations (Nolan 2011: 9–14).
7) Counterfactuals are needed in causal explanations: we take those factors to be
explanatory without which the event to be explained would not have occurred
(Nolan 2011: 14–15).
8) Counterfactual information is needed in assessments of moral responsibility: we
need counterfactual information in order to appraise or blame historical actors.
The attributions of moral responsibility partly depend on considerations whether
an outcome that an actor is blamed or appraised for would have happened
without the actor’s intentional involvement. (Nolan 2011: 15–17).

I will not assess whether Nolan is correct in all of his claims or whether his
classification is correct or not, since it is not crucial for the task at hand in this thesis.
What is of interest is Nolan’s treatment of the uses 6) and 7). Many writers who link
counterfactuals to the issues of causation, causality, causal reasoning or causal
explanation (Fogel 1964; Hawthorn 1991, cp. 2; Fearon 1991; Tetlock & Belkin 1996;
Stanford 1998: 85–93; Bulhof 1999; Lebow 2007; Megill 2007: 100, 155; Kaye 2010),
argue for their position in a way that seems quite loose from the point of view of
someone who seriously tries to see whether the proponents are right in their claims.
Nolan’s arguments are no different in this respect, and can therefore be used to point of
the general need for more substantial argumentation on the relation between causal
claims and counterfactual claims.

2.3.2!Mysterious!link!between!causal!explanatory!claims!and!counterfactual!claims!
When Nolan discusses what he calls “causation” and causal explanations his aim is to
conclude that there are reasons to suggest that there is a connection between causal
reasoning and counterfactuals. In the end, however, Nolan does not seem to find good

17
reasons for this conclusion, and therefore ends up occupying quite vague stance on the
question. First, Nolan mentions that it is controversial among philosophers what the
connection between causal claims and counterfactuals precisely is (Nolan 2011: 10).
Then he notes that “there are arguably several other ways to have causation without this
sort of direct counterfactual dependence” (ibid.), but this claim is not presented with any
specific examples or argumentation. The only philosophical theory of causation that
Nolan mentions is David Lewis’s account on counterfactuals and causation, but the
conclusion is gloomy: due to the several problems that the account has been shown to
face counterfactual analyses are not very popular among philosophers these days (Nolan
2011: 10–11). In spite of these problems that Nolan presents, he still wants to conclude:
“the close links between causal and counterfactual judgments mean that historians
interested in causation should not reject a role for counterfactuals in historical
reasoning” (Nolan 2011: 13). The conclusion is that there seems to be a connection
between counterfactuals and causation, but the reasons for why there is such a
connection are not spelled out.
The same kind of looseness is apparent when Nolan discusses the role of
counterfactuals in causal explanation. Nolan mentions James Woodward’s (2003) view
that explanations are answers to counterfactual what-if-things-had-been-different
questions, but he does not pursue the idea any further (Nolan 2011:14). Instead, Nolan
continues by claiming that although counterfactual theories of causal explanation might
be mistaken (ibid.), “explanations which specify conditions without which the
phenomenon to be explained would not have happened seem to be of particular interest”
(Nolan 2011: 15). Counterfactuals are deemed as important, but the justification for
why they are important is based on a few examples. For the one who doubts the
methodological soundness of counterfactual reasoning this might seem far from
convincing once again.
My purpose is not to claim that Nolan fails in his argument for the
indispensability of counterfactuals altogether, because he is not shedding much light on
the connection between causal and explanatory claims and counterfactuals. Nolan’s
purpose is to show that counterfactuals are ubiquitous in historiography and that there
are reasons to think that counterfactuals are methodologically useful in various cases. In
this he succeeds. My point is to exemplify the lack of theoretical justification that one
comes across when reading through the protagonists’ arguments. The idea that there is a

18
link between counterfactual reasoning and causal reasoning is often brought up, but the
idea is not extensively argued for.
In general, it seems that quite often the argument for the connection rests on
nothing more but on our everyday tendency to think of causation in counterfactual
terms. In addition, in cases where more explicit theoretical ideas of the relation between
counterfactuals and causal claims are presented, they are either briefly discussed or just
mentioned by referring to unnamed philosophers who think that the counterfactual
analysis of causal claims is the right way to go (e.g. Megill 2007: 100). The
protagonists are, hence, in need for a philosophical theory of counterfactuals that would
turn their idea of the connection between counterfactuals and causation into a serious
argument that can be used to defend their position.

2.3.3!Causation,!causal!relations!and!causal!explanation!
The arguments of the protagonists of counterfactuals contain also another possible
source of confusion: the terms causation, causality, causal claims, causal
reasoning/thinking and (causal) explanations are used in different ways in different
contexts. For instance, counterfactual thinking is sometimes claimed to be relevant for
causation whereas sometimes it is presented as important for causal explanation.
Sometimes these terms are used interchangeably without paying any special notice to
this, whereas sometimes the issues of causation and explanation are discussed
separately like in Nolan’s paper. Based on the way that these concepts are used in
philosophy, the separation is well justified. The terms causation and causality are
usually used to denote the causal relations in the world, and the term causation is
especially used in the context of metaphysics when discussing the nature of these
relations.
The true nature of causal relations and the issues of causation are not, however,
the issues that concern (or should concern) historians the most. Historians’ interest lies
in how to answer questions ‘what caused x?’ and ‘why x?’ and the claims ‘y caused x’
and ‘y explains x’ that are used to answer these questions. The point of the questions is
to find the factors that account for x. The point of the answers is to present at least one
factor y that caused and therefore explains x. Historians are therefore interested in i) the
specific causal relation between y and x, ii) the causal and explanatory claims that
represent the relation and iii) the explanatory and causal reasoning that is involved in
the construction and justification of these claims. The question of counterfactuals

19
therefore rises up in this realm of causal explanation: what is the relation between
counterfactual reasoning and causal explanatory reasoning? Is counterfactual reasoning
merely an apt way to approach causal claims and causal reasoning in some cases, or is
there actually a conceptual relation between causal claims and counterfactual claims
that automatically links all causal reasoning to counterfactual reasoning?
In order to avoid conceptual confusions, I will adopt the following way of using
causal terms in this thesis; causation and causality are terms used to refer to
philosophical views on the nature of causal relations. The term causal relation is used
when speaking of specific causal relations that historians suggest in their explanatory
and causal claims of the general form ‘y explains x’. The task at hand is therefore to
present a theory of causal explanation – not of causation or causality – that can serve to
further elaborate the link between causal explanatory claims and counterfactual claims.
In this context, when the word cause is used it means an explanatory cause – a factor
that is used to explain certain event x. Now that the terminology is in place, it is time to
further clarify the protagonists’ argument.

2.3.4! Argument! for! the! use! of! counterfactuals! in! historiographical! causal!
explanations!
The argument of the protagonists of counterfactuals who claim that counterfactuals are
methodologically justified in the context of causal explanations can be reconstructed as
follows:

1) The task of the historian: Historians do not merely describe or recount past
events, they also try to account for or explain them.
2) The nature of explanation in history: Since explanation in history is causal
explanation and the causal explanation of events is the business of identifying
causes of events, historians are in the business of making and justifying causal
explanatory claims.
3) The nature of causal claims: Causal explanatory claims have an inherent
conceptual connection to counterfactual claims and therefore every causal claim
carriers with it a counterfactual commitment.

The antagonists and protagonists of counterfactuals do not necessarily disagree on the

first and second premise. E.H. Carr, despite of his despise for counterfactuals, thought

20
that it is the task of historian to explain events by specifying their causes (Carr 1964:
105). So, the disagreements most likely concern the third premise: the antagonists and
protagonists disagree on the conceptual issue whether causal claims are inherently
related to counterfactual claims. As already pointed out, this is the premise that the
protagonists of counterfactuals have not paid too much attention to, although it is
central to their main claim that counterfactuals are indispensable for a historian.
The third premise says that there is a conceptual relation between counterfactual
claims and causal explanatory claims. In the context of confirmation of hypotheses this
means that once one has justified a causal claim one has justified the related
counterfactual as well. This claim does not imply any commitment to a stronger,
according to which one can only justify a causal claim by first justifying the
counterfactual claim with counterfactual inferences. Although the volumes edited by
Tetlock & Belkin (1996) and Tetlock, Lebow and Parker (2006), are dedicated to
counterfactual thought experiments that are purported to support specific causal claims,
these authors seem not to be committed to the stronger thesis; the are only committed to
idea of conceptual relation (Tetlock & Belkin 1996: 3–4; Tetlock & Parker 2006: 17–
18).
The conclusion of the chapter is that there is a need for a philosophical theory of
causal explanation that explicitly shows how causal claims are related to counterfactual
claims. This theory will have an explicative justificatory role from the point of view of
the protagonists of counterfactuals. The purpose of this thesis is, however, to go beyond
just this mere explicative role and to provide an account of causal explanation and
counterfactuals that provides some methodological fruits as well. Showing that – and
explaining why – one can methodologically gain something by using counterfactuals
further justifies the use of counterfactuals. The following chapters of the thesis are will
show that the contrastive counterfactual theory of causal explanation is the theory that
fulfills these desiderata.

21
3! !Contrastive!counterfactual!theory!of!causal!explanation!
!!!!!!!
Counterfactuals, causes, and explanations are three sides of the same strange three-
sided coin; you cannot have one without the other two.
– Johannes Bulhof (1999: 147)

The philosophical theory of causal explanations and causal claims I use to pin down the
role that counterfactuals play in historiographic explanations can be called
counterfactual theory of causal explanation (Woodward 2003) or contrastive
counterfactual model or account of explanation (Ylikoski 2001: 7; 2007). Although
Woodward’s and Ylikoski’s accounts of causal explanation are not similar in every
detail1, the accounts share sufficiently similar core ideas. Thus, I see myself entitled to
frame them as proponents of the same type of theory, and to use the ideas and
arguments of both of them for the purposes of developing my own argument. The
contrastive counterfactual theory combines the idea of contrastive explanandum with
some of the insights that have emerged in the attempts to analyze causal claims and
causation2 in counterfactual terms. The purpose of this chapter is to present the main
theses of the account. First, I will briefly present the background of the counterfactual
contrastive theory, and the ways in which the ideas central ideas of this theory have
been put to use in philosophy of historiography in recent decades. Then I will present
the main theses of the account in a more systematic manner.

1
For instance, Woodward’s theory is extremely rich in technical definitions and thoroughness while
Ylikoski’s account is more accessible in its argumentative clarity. Woodward is on the quest of
establishing a theory of causation while Ylikoski’s interest lies in the establishing a theory of causal
explanation that is not depended on what kind of theory of causation we adopt or end up adopting – be it
counterfactual, probabilistic, unificationist or something else (Ylikoski 2001: 11). In addition, Ylikoski
(2001: 8) applies his view to singular causal explanations, whereas Woodward aims at providing a
general theory of causal explanation that applies both to generalizations and to singular facts as
explananda (Woodward 2003: 17).
2
One can distinguish between analyses of causal explanatory claims and analyses of causation as a
relation in the world as Ylikoski (2001: 11) does. Analyses of causal explanation are accounts on how
causal explanation works and what kind of features causal explanations have, whereas the analyses of
causation deal with the metaphysical question on the nature of the causal relations in the world. My focus
is on how causal and explanatory inference works in historiography. As to the question of causation, I
will adopt Woodward’s (2003: ch.6) idea of invariance as a key feature of causal relations, which suffices
for the purposes of this thesis.

22
3.1!Two!paths!to!the!contrastive!counterfactual!theory!

3.1.1!Counterfactual!theories!of!causal!explanation!
The point that causal claims are cashed out in counterfactual terms is not a novel one.
As we saw in the previous chapter, the defenders of counterfactual historiography have
a strong more or less intuitive idea that causal claims and counterfactual claims hang
together in some way. Counterfactual interpretations of causal claims have been around
as long as causation and causal claims have been analyzed. David Hume, who has not
usually been regarded as a strong defender of counterfactual theories of causal claims,
does not seem to have been able to resist the quite natural inclination to interpret causal
claims with the help of counterfactuals: “We may define a cause to be an object
followed by another, and where all the objects, similar to the first, are followed by
objects similar to the second. Or, in other words, where, if the first object had not been,
the second never had existed.” (Hume 1748: Section VII.)
As already pointed out in the previous chapter, David Lewis (1973a; 1973b;
1986b; 2000) has been the main proponent of the counterfactual theory of causation and
causal explanation in recent decades3. Despite Lewis’s influence in the field of
metaphysics of causation, his theory has not been put much to use in philosophy of
science. In philosophy of historiography Lewis’s views have been only briefly referred
to in the contexts defending the use of counterfactuals in causal explanations (e.g:
Bulhof 1999: 167; Climo & Howells 1976; Fearon 1991: 176; Tetlock & Belkin 1996:
18, Nolan 2011: 10). The most extensive discussion and criticism of Lewis in the
context of historiography can be found in Hawthorn’s Plausible Worlds (1991: 21–25)
and in Elster (1978: ch. 6).
The quite minor impact that Lewis’s theory has had can be contrasted with
Woodward’s impact in philosophy of special sciences in recent years. The fact that
Woodward’s theory seems to solve most of the conceptual problems that have bugged
Lewis’s account is a merit in itself and of high interest to a philosopher4, but much of
the charm of Woodward’s view is derived from its close connections to work done in

3
Also Horgan (1989) and Ruben (1994) have proposed versions of counterfactual theory of causal
explanation in recent decades.
4
In fact, in his paper on ”Causation as Influence” (2000), Lewis presents the new version of his
counterfactual theory in order to avoid the problems that have haunted his earlier account. This new
version comes rather close to the main ideas of the contrastive counterfactual theory. In this paper, Lewis
defines that causation in terms of ”how-how” counterfactual dependence: if a had happened in this way,
then b would have happened in that way. In other words, the ways in which b would happen are
systematically dependent on the ways in which a happens.

23
statistics and econometrics. Woodward employs the means that, for instance,
econometricians have used to distinguish causal relations from mere correlations and
gives these ideas a more general and a more philosophical form. In this way, Woodward
manages to present a theory that seems to be applicable way beyond the field of
econometrics5.
Besides the general ideas presented by Raatikainen (2011), there have not been
any systematic attempts to see whether Woodward is able to deliver the goods for a
historian. Thus, this thesis works as a test bed for the contrastive counterfactual theory
in historiography. If the test succeeds, there is a case to claim that the explanatory
inferences of historians do not substantially differ from the explanatory practices of
other fields although there are surely other kinds of questions and methodological
challenges that are more unique to historiography. If the test fails, the antagonists of
counterfactuals have a new ace on their sleeves: counterfactuals do not seem to play a
significant role in historians’ explanatory and causal reasoning.

3.1.2!Contrastive!approach!
The idea of contrastive explanandum has gained popularity in philosophy of science in
recent decades. According to the proponents of the contrastive view of explanation,
explanations are answers to questions of the form of ‘why did f occur instead of c’ (see.
e.g. Garfinkel 1981: 21-48; Lipton1990; Ylikoski 2001: 8). The term f stands for any
property, event, action or any other thing the occurrence of which one is interested in
explaining. The contrast of f is c – a property, event or action that cannot occur together
with f. In other words, f and c are mutually exclusive contrasts.
The contrastive account argues that explanation is not about plain facts but about
contrastive facts of the type ‘f rather than6 c’. Alan Garfinkel (1981: 21–22) introduces
the old joke about the bank robber Willie Sutton as an example of contrastive
explanandum. A priest went to see Willie Sutton when Willie served time in jail. The
priest inquired politely why Willie robbed banks. Willie answered bluntly: “Well, that’s
where the money is”. The joke is based on the different contrasts that the priest’s
question and Willie’s answer have. The priest’s question seeks an answer to the contrast

5
For instance, Woodward’s theory has been applied to ecology (Raerinne 2011), equilibrium
explanations in economics and ecology (Kuorikoski 2007). Woodward’s (2003) own examples taken
from the sciences cover e.g. mechanics (12–14), genetics (302–307), and medicine (312–313).
6
There are several ways to state the contrast of an explanation-seeking question in different languages
(van Fraassen 1980; Garfinkel 1981: 25, 29; Ylikoski 2001: 24). For instance, in English one can ask:
“Why f rather than c?”, “Why f and not c?”, “Why f instead of c?”.

24
“why do you rob banks instead of being honest and law-abinding?”, whereas Willie
answers the question “why to rob banks instead of grocery stores or pharmacies?“.
As the above example shows, the advantage of using an explicit contrastive
explanandum is that a contrastive explanandum clarifies what the explanatory-seeking
question is about. When the explanandum is stated in contrastive terms, it is also easy to
judge whether explanations that seem to be competing actually are competing or not. If
explanations answer the same contrastive question, they are competing, but if they give
answers to slightly differing contrastive questions they are complementing rather than
competing explanations (Ylikoski 2001: 22).
In the context of philosophy of historiography, the gains of using contrastive
explanandum have not been left unnoticed. Jerouen van Bouvel and Erik Weber (2008)
have used the contrastive approach to defend explanatory pluralism7. Raymond Martin
(1989: 59–65), James Fearon (1991: 172; 1996: 56–59), Aviezer Tucker (1998; 2004:
192–194), Tim de Mey and Eric Weber (2003) all introduce the idea of contrastive
explanandum in order to tackle various problems related to philosophy of
historiography8. None of these writers, however, links the idea of contrastive
explanandum to a counterfactual theory of causal explanation in the way I will present
next.

3.2!Main!ideas!of!contrastive!counterfactual!account!
The core theses of the contrastive counterfactual theory of causal explanation can be
characterized as follows:

1) Explanations are answers to questions. The questions can be formulated in

several ways: they can e.g. be why-questions or mechanistic how-questions
(Ylikoski 2001: 21).
2) All adequate causal explanations have a counterfactual form (Woodward 1984:
236–245; Woodward 2003: ch. 5; Ylikoski 2001: 8).

7
The pluralist position maintains that there are no a priori reasons to prefer one type of explanation – be it
intentional, structural, “micro”, “macro”, functional etc. – to any other type of explanation. According to
van Bouvel and Weber (2008), the contrastive explanandum helps to define what type of explanation
might work best in answering the given contrastive explanatory question.
8
Martin tries to defend the view that weighted explanations in historiography can be objective; Fearon
takes note that the use of contrastive explanandum can facilitate hypothesis testing in international
relations, and that failures in specifying the contrast explicitly can, on the other hand, lead to problems in
testing; Tucker tries to show that historians who disagree with each other are usually trying to answer
different explanatory questions; de May and Weber tackle the question of causal weighting by using
contrastive explananda and counterfactuals.

25
 3) All explananda are best conceptualized as contrastive (Woodward 1984: 245 –
249; Woodward 2003: 145–146, 212–213; Ylikoski 2001: 8, 31).
4) Explanations track objective relations of dependency in the world (Woodward
2003: 118–121; Ylikoski 2001: 19). These objective relations of dependency can
be characterized as invariant under some interventions (Woodward 2003: ch. 6).

My intention is not to argue extensively for these theses since it is done in a detailed
way both by Woodward and Ylikoski. The first thesis spells out the main idea of
erotetic approach developed by van Fraassen (1980), Tuomela (1980), Achinstein
(1983) and Sintonen (1984): explanations are answers to questions and the research
proceeds by answering and elaborating questions. What the contrastive counterfactual
account adds to this is the idea that explanatory reasoning works with contrastive
explanation-seeking questions. We try to find those factors that account for the
difference stated in the contrast. In this we engage in counterfactual reasoning: If these
factors had been different in some respect(s), the contrast c would have occurred instead
of the fact f that actually occurred.
One consequence of the contrastive idea is the claim that when we are
explaining a contrast we are explaining some aspects of a phenomenon or an event or a
series of events. Therefore we are not seeking to explain the whole phenomenon, event
or series of events (Ylikoski 2001: 20). This means that we can, for instance, explain the
occurrence or the non-occurrence of an event or particular properties of it, but not all
the properties at once (Woodward 1984: 247–249). This is because the contrast always
picks one aspect to be explained: we can ask why the ball is blue instead of pink or then
we can ask why the ball is colored instead of it being colorless. These are different
questions concerning different aspects of the ball, and therefore, they presumably have
different kinds of answers.
The contrastive idea is not a generalized descriptive statement of all the
explanatory seeking questions in the world. No one is claiming that all questions in fact
have an explicit contrastive form. We do not usually state the contrast of our
explanation seeking questions explicitly, because the contrast of our question is often
obvious in the context where the question is presented. What the contrastive idea is
claiming is that all explanation seeking question have an implicit contrastive structure
that can be spelled out explicitly when needed, for instance, in cases where there is
confusion about what purported explanations actually can explain.

26
3.2.1!Contrastive!counterfactual!account!of!singular!causal!explanation!
The contrastive counterfactual theory of causal explanation can be applied both to type
causal claims and to singular causal claims. Type causal claims say in generic terms
how certain factors of type A (e.g. smoking) cause changes in other factors of type B
(e.g. chronic bronchitis). Singular causal explanations, on the other hand, involve token
causal claims that state the causes of some particular event, action or series of events.
Since historians rarely explain generic repetitive patterns, repetitive phenomena or
generalizations, I will limit my focus here on the characterization of the role of
counterfactuals in singular causal explanation and the analysis of token causal
explanatory claims. The central task is to shed light on the ways in which counterfactual
reasoning works when answering questions like ‘why did this certain event occur?’.
The contrastive counterfactual theory analyzes singular or token causal claims of
the form ‘a caused b’ as follows:

(3.1) a [a*] caused b [b*]

or interchangeably9
(3.2) a [a*] explains b [b*],

In this formulation, a is a particular event or action or a set of events or actions [a1…an]

that is claimed to have caused, and therefore to explain, a certain particular event or
action b, whereas a* and b* are the contrasts of a and b. In Woodward’s terms, a and b
are the actual values of two variables A and B. a* and b*, on the other hand are
alternative values to a and b that the variables A and B could have taken, but in the
actual case did not take10. Variables are properties (or magnitudes) that can take more
than one value (Woodward 2003: 39). For instance we can have the variable of color C
that can have several different values [c1 = red, c2 = blue, c3 = orange, c4 = pink etc.].
The advantage of using the notion ‘variable’ also in singular causal explanation is the
fact that it is an apt way to refer to all the various terms like ‘event’, ‘action’, ‘process’,

9
As I said in the chapter 2 (section 2.2.3) I use the term cause to refer to explanatory causes. That is why
the verbs to cause and to explain are used interchangebly.
10
I will mark variables with capital letters such as A, B, X, Y and the values that these variables take with
a, b, x, y and the contrastive values with a*, b, x*, y*.

27
‘property’ or ‘phenomenon’ that one uses in one’s explanation. I will also speak of a
and a* as facts11.
The main crux of this analysis is that singular causal explanations ask why a
certain variable took the value a it actually did instead of some other value it could have
had (why this fact a instead of its contrast a*?). For instance we can ask why the ball is
blue instead of being pink. In addition, we could also ask why the ball has a color blue
rather than it being colorless, in order to inquire why a certain variable had a value in
the first place. The point is that specifying the contrast(s) of the question “why is this
ball blue?” helps to clarify what we are in fact asking: are we interested in knowing why
the ball has just the color it has or rather why it has a color in the first place?

3.2.2!From!causal!claims!to!counterfactuals!
The idea of contrastive explanandum has been widely promoted in the literature of
causal explanation. However, it has not been too common, to state the explanans or the
token cause in contrastive terms a[a*] as I have done here following Jonathan Schaffer
(2005) and Petri Ylikoski (2011). The reason why I have now chosen to present the
singular causal claims with double contrasts, is the fact that the formulation shows
explicitly the structure of the counterfactual claims that are implied by token causal
claims. Put in more exact terms, the claim ‘a[a*] explains b [b*]’ implies the following
two counterfactuals:

(3.3) if a had not happened, b (defined as a contrast to b*) would not have
happened
(3.4) if a* had happened, b* would have happened

For instance, in the case of the contrastive question why the ball is blue rather than pink,
(which was the original color of the ball), we can say that it is blue because Mary
painted it blue. The contrasts of the explanandum are [b = the ball is blue; b* = the ball
is pink] and the contrasts of the cause that explains the asked contrast are [a = Mary
painted it blue; a* = Mary did not paint it at all]. The token causal claim implies two

11
When a variable A takes a certain value a or a*, this constitutes a fact that a or a fact that a*.

28
counterfactual claims. 1) If Mary had not painted the ball blue, the ball would not be
blue. 2) If Mary had not painted the ball blue, the ball would be pink.
The main plot of the contrastive counterfactual theory runs therefore as follows:
in singular causal explanation, we explain the contrast of the explanandum by
specifying the factor or group of factors that account for the asked difference specified
in the contrast. The causal claim states that there is a causal dependency relation
between the explanans and explanandum and the dependency relation can be
characterized in counterfactual terms. The formulations (3.3) and (3.4) are simple ways
to introduce the basic idea of the counterfactual analysis of causal claims, but they do
convey the idea of intervention that is central for making the counterfactual analysis to
work. It is therefore time to further characterizing the semantic workings of
counterfactual claims and to discuss the concept of intervention and the more precise
formulation of dependency relation.

3.2.3!Tracking!invariant!dependency!relations!with!interventions!
Following Woodward’s (2003) manipulationist account of causation the truth
conditions for singular causal and explanatory claims are characterized as follows:

(3.5) a[a*] causes b[b*] if and only if there is an intervention on the variable A such
that: if the intervention changed a to a*, this would bring about a change from b
to b*.

What does it mean to intervene on a[a*]? An intervention on some value x of the

explanans variable X with respect to the explanandum variable Y is a causal process
that changes the value of X in an appropriately exogenous way, so that if a change in
the value of Y occurs, it only occurs in the virtue of the change in the value of X
(Woodward 2003: 94). Interventions are also called ideally surgical manipulations (e.g,
Kuorikoski 2010: 23) that affect only the explanans variable and leave the values of all
other variables intact and show what kind of effect the manipulation of the explanans
variable has on the explanandum variable.
Since interventions are characterized with the help of other causal concepts, the
characterization (3.5) cannot provide a reductive analysis of the concept of cause.
Another interesting point is what the characterization (3.5) suggests about “the origins”
of our causal reasoning: the way in which we humans infer and reason about causal

29
relations has its roots in our ability to act as manipulating agents who actively intervene
on the workings of the world. We want knowledge of the world that helps us to
manipulate our surroundings, that is, causal knowledge. Our beliefs and ways of
reasoning, however, do not determine the truth of the causal claims that are of interest to
us: it does not depend on the beliefs and reasoning of the agents whether bringing about
a[a*] brings about b[b*]. The condition of a successful manipulation is therefore that
there exists a causal dependency relation that our causal claims and the related
counterfactuals manage to capture. Due to this, the contrastive counterfactual theory is
committed to a realist stance (Woodward 2003: 118–119).
Another important point of Woodward’s manipulationist account is that its
applicability is not limited to those situations in which we are actually in a position to
manipulate certain variables. This is good news for the disciplines such as
historiography, various social sciences, ecology, or cosmology whose subject matters
hardly provide any opportunity for straightforward experimental manipulations. The
wide applicability of Woodward’s account stems from his idea of intervention that is
not in any special way connected to human action. An intervention can be any causal
process that changes the value of X and controls the values of other variables Z1 … Z2
so that X becomes the only variable that the values of Y are dependent on. Thus, events
of nature that occur without any human involvement as well as experimental
manipulations can both be regarded as interventions if they work in the way that
interventions should: they turn the variable X into “a switch” that controls the values
taken by the variable Y.
The idea of intervention also gives the criteria to distinguish real causal relations
from correlations and other relations that only appear to be causal: mere correlations
cannot be used to successfully manipulate b[b*]. Woodward (2003: ch. 6) calls the
property that makes successful manipulation possible invariance under interventions.
For a relation to pass as invariant there must be at least one intervention on the
explanans variable that changes the value of the explanandum variable in the way that
the causal claim says it will. Mere correlations do not pass this test, whereas real causal
relations do. In this way Woodward’s account solves one of the key problems that has
haunted several other analyses of causation and causal explanation.
The well-known example of a barometer reading predicting the weather but not
explaining it serves as an example that nicely illustrates the idea of intervention. We can
manipulate the values of the barometer but not one of these manipulations seems to

30
have any systematic effect on the weather – changing the barometer reading from 1040
mbar to 950 mbar does not make fair weather turn into a storm and one should not
expect a nice weather for an outdoor day after tempering with the readings of one’s
barometer. The relation between the barometer readings and the weather shows no
invariance. There are no interventions on the barometer readings that would change the
weather. The relation is a mere correlation.
Instead, the relation between air pressure and weather exhibits a certain degree
of invariance. There are interventions on the air pressure that if they were carried out in
some way they would change the weather. For instance, if there was an intervention that
would change the high pressure to extreme low pressure, there would be a change from
fair weather to rain or storm. So, the variable whose changes explain the contrast
between fair weather and rainy weather is the air pressure, not the barometer readings
for which the air pressure also accounts.
The above example also helps to illustrate the feature that distinguishes
Woodward’s account from standard naïve analyses of causation in counterfactual terms:
the account saves the asymmetry of causal relations and does not allow backtracking
counterfactuals. The naïve counterfactual reasoning that follows the lines of the
characterizations (3.3) and (3.4) easily leads to results where the cause could be deemed
as counterfactually dependent on the effect. In other words, if we took counterfactuals
like (3.3) and (3.4) to be sufficient to determine causal relations, we would have causal
relations where the effect would be the cause of the cause! For instance, we could infer
that barometer readings are the cause of the low pressure with the following inference:
if the barometer readings had been low, there would certainly have been a low pressure.
This example brings out the well-known point that not all counterfactual
dependencies correspond to causal or explanatory relationships. Epistemically, it is
justifiable to infer the occurrence of storms based on the barometer readings, but these
inferences do not reflect causal relations in the world. Woodward’s notion of
intervention is, therefore, used to distinguish the true causal relations and the related
counterfactuals from the counterfactual inferences that do not correspond to causal
relations.
As already mentioned, in the case of singular causal claims and singular causal
explanations causal explanatory claims can be characterized in interventionist terms as
follows:

31
 a[a*] causes b[b*] if and only if there is an intervention on the variable A such that:
if the intervention changed a to a*, this would bring about a change from b to b*.

This condition clause states that there should be an invariant relation between a[a*] and
b[b*] if the relation is to be counted as causal. Thus, in order for a causal claim to be
explanatory it should support the right sort of counterfactuals about what would happen
under interventions. So, Woodward’s manipulationist account does not merely say that
causal claims should support any counterfactuals but rather that they should support
counterfactuals formulated in interventionist terms. I call these counterfactuals that are
needed in causal explanation interventionist counterfactuals. By making use of
interventionist counterfactuals we can track invariant dependence relations in the world,
which are, the truth-makers of causal explanatory claims. Interventions and invariant
dependency relations will be more thoroughly discussed in chapter 5 that will focus on
the epistemology and methodology of counterfactual inferences in historiography. Next,
however, we will move on to examine how the general framework and conceptual tools
presented in this chapter solve some of the central problems of historiographical
explanation.

32
4!!
!!!!!! !
Contrastive!counterfactual!theory!in!historiography!
!
‘If there is such a thing as a sense of reality’, Robert Musil had agreed in The Man
Without Qualities, … ‘– then there must also be something that one can call the sense of
possibility’. But in history and social sciences it has not been usually clear what there is
to be gained in developing it.
– Geoffrey Hawthorn (1991: 4)

When browsing through the literature on explanation in historiography, one can’t help
getting the impression that there are as many philosophies and methodologies of
historiography as there are philosophers and historians writing on them. The purpose of
this chapter is to lay the basis on a coherent and unified view of causal explanation in
historiography based on the contrastive counterfactual theory introduced in the previous
chapter. Various themes ranging from the problems of explanatory selection, causal
background and negative causes to the problem objectivity and subjectivity are
discussed in order to show how they can all be fruitfully handled with the idea of
contrastive explanandum and contrastive causal claims.
In the end of the chapter, the contrastive counterfactual model is compared to
the best known, although conflicting, philosophical accounts of historiographical
explanation: the covering law model and the non-causalist or intentionalist view of
explaining actions. The covering law model is shown to face some serious problems,
and therefore, not to be expected to provide fruitful explanatory methodology for
historiographical explanation in the same way that the contrastive counterfactual model
provides. The non-causalist position is answered by arguing that reasons can be treated
as causes in the framework of the contrastive counterfactual theory without
undermining their nature as reasons.
The overall argument of the chapter is the following: on the basis of the
discussions in section 4.1 and on the basis of the comparisons to the best-known
alternative theories of historiographical explanation in section 4.2 the contrastive
counterfactual theory can be regarded as the most fruitful theory to characterize the
nature and problems of historiographical explanation.

33
4.1! Contrastive! explananda( and! contrastive! causal! claims! in!
historiography!

4.1.1!The!contrastive!explanation!of!facts!
It is necessary to start with specifying what historians are explaining in their
explanations. The general term I use to denote the target of explanation is fact. Facts are
events, series of events, processes or phenomena that historians are interested in1. In
woodwardian terms facts are constituted by variables taking a certain values. I will call
historiographical claims that refer to facts that have happened in the past,
historiographical descriptions. As already pointed out in chapter 2, historiographical
descriptions are claims of the facts of the past and they are the end result of
historiographical research; historians infer what happened in the past with the help of
the traces of that past salvaged to this day (Tucker 2004: 17, Collingwood 1994[1946]:
251–252).
As also pointed out in chapter 2, an interesting feature of historiographical study
is that the problem of underdetermination is often faced also in practice due to lack of
evidence. The lack of evidence is one reason why historians often disagree on purported
facts without reaching an agreement. For instance, the gaps in data make it a detective
story to find out whether there were peasant revolts in the Classic Period Lowland Maya
civilization as J.E.S Thompson (1966) suggests. Some facts have, of course, strong
evidential support and most historians agree on them. For instance, the fact that John F.
Kennedy became the 35th President of the United States on January 20th 1961 is hardly
under dispute.
In cases where historians have reached an agreement on what happened, they
might still argue on what to call the things that happened. Indeed, the most famous
historiographical disputes have centered on the names given to the long term processes
such as “the First Industrial Revolution” or “the fall of the Roman Empire”. In these
cases, the debate has mainly been on whether the terms “revolution” or “fall” capture
the nature of the changes that are inferred from the data. One task of philosophy of
historiography would then be to explore how these kind of descriptive disputes develop

1
There is a general discussion within philosophy of science whether facts should be considered as the
right relata for causal explanation (see e.g. Mackie 1974: ch. 10 and Ruben 1990: ch. 5). In the
metaphysics of causation, the question has been whether the relata of causation are facts (see e.g Mellor
1995). I do not have the opportunity to pursue these questions any further in this thesis where the central
question is causal explanation in historiography, not causal explanation in general.

34
and how they are linked to explanatory competition2. Unfortunately, in order to keep my
lines of argument together, I do not have the possibility to explore this question here.
Another important issue I cannot discuss in detail is the question how counterfactual
and causal reasoning might work together when assessing whether the pieces of
evidence, which one has, provide reliable evidence for a certain fact or not (see section
2.2.3 however). Thus, the explanations that are the focus of my thesis are the
explanations of the facts that historians have inferred from the data and which they
mainly agree on.
The starting point in applying the contrastive counterfactual theory to
historiography is the idea that historians explain facts - events, series of events,
processes or phenomena of the past – and that these facts should be conceptualized in
contrastive terms3. Take, for instance, the presidency of John F. Kennedy. We can ask
why John. F. Kennedy won the Presidential election of 1960, and implicitly assume the
following contrastive question: why did John. F. Kennedy win the Presidential election
of 1960 instead of Richard Nixon? The explanatory task for the political historian would
then be to try to pick out the factors that caused the fact that it was JFK who gave the
inaugural address on January 20th 1961 instead of Nixon. This explanatory information
would ideally be of the sort that it would allow the historian to answer what if things
had been different-questions such as “what if Kennedy had performed as badly as Nixon
did in the first televised U.S. presidential debates? Would he have been elected then?”.
This example shows how the idea of contrastive explanandum leads to counterfactual
considerations.
My point here is that it is not just the cases where the contrast seems fairly
obvious that should be interpreted in contrastive terms. For instance, the classical
question “Why did the Greeks win the Battle of Salamis in 480 BCE?” can be thought
to have several contrasts. The most obvious contrast in this case is: “Why did the
Greeks win the Battle of Salamis instead of the Persians?”. But there are also other
contrasts that intrigue the historians such as: “Why did the Greeks win the Battle of
Salamis and not the previous sea battles?” or “Why did the Greeks win the Battle of
Salamis quite overwhelmingly and not just barely?”. It is interesting to note that a

2
Raymond Martin (1989: ch.3) has taken a step into this direction with the case study on the disputes of
archeologists and historians on the population loss of the Classic Period Lowland Maya civilization
3
The natural idea is that explanations cannot explain the things of the past in themselves. The past events
need be conceptualized in some way to be a target of explanation (see Ylikoski 2001: 20). The idea of
contrastive explanadum is a way to conceptualize facts for explanatory purposes.

35
simple-looking question like “Why did the Greeks win the Battle of Salamis?” can have
various contrasts, and due to this different meanings, but the point goes deeper here.
What is worth noticing is the fact that these different contrastive questions usually
require somewhat different kinds of answers. The contrast thus determines what kind of
factors one should include in one’s explanation of the contrast – and it might quite often
be that these are not the same factors in all the contrasts.
There are numerous examples of contrastive questions; indeed, if the hypothesis
of the proponents of the contrastive approach is correct there are at least as many
examples than there are causal explanatory-seeking questions. This is why I will settle
for one more: the process called the First Industrial Revolution. The point of this
example is to make clear “the corollary” of the contrastive idea – the contrastive idea
implies that we are explaining aspects of events or series of events, not the events or
series of events as a whole. The reader might have noticed this with the previous
examples, but with a long-time process consisting of thousands of events like the First
Industrial Revolution one cannot expect to be able to provide an all-compassing
narrative of the things that happened and link them causally together as a total complete
explanation. What some economic historians have done instead is that they have
explained the process of British industrialization through contrastive questions. They
have asked: “Why did Britain industrialize instead of other countries in continental
Europe?” (Crouzet 2001: 111: 114), “Why did the cotton industry in Britain become
factory based instead of India taking the lead?” (Landes 1998: 225–30), and “Why
Western Europe and not China?” (Pomeranz 2006).

4.1.2!Explanatory!selection!in!fixing!the!causal!background!!
What would be wrong in trying to explain an event as a whole? In principle, nothing if
we take the explanation of the whole event to mean that we would be able to explain all
the contrasts of all the aspects of an event (Ylikoski 2001: 20). In practice this, of
course, might turn out to be an insurmountable task. Consider, for instance, what it
would mean to explain all the contrasts of all the aspects (like cotton industry, mining,
organization of labor, the institutional context) of the First Industrial Revolution. Carl
Hempel presents the problem of complete or total explanations nicely as follows:

36
 What is sometimes called the complete description of an individual event
(such as the earthquake of San Francisco in 1906 or the assassination of
Julius Caesar) would require a statement of all the properties exhibited by
the spatial region or the individual object involved, for the period of time
occupied by the event in question. Such a task can never be completely
accomplished.
(Hempel 1942: 37.)

Peter Railton’s (1980) idea of ideal explanatory text is one way of defining what it
would mean to explain events as a whole. According to Railton ideal explanatory text is
a theoretical concept that includes all accurate information about every due-to relation
relevant for the explanandum (Railton 1980: 147-148). All information that reduces
insecurity of the things that account for explanandum is deemed as explanatory. Førland
(2004) therefore praises Railton’s account since it “invites the historian to search for
ever more explanatory information”. Seeking explanatory information is naturally a
good thing, but thinking in Railton’s terms leads to the problem of explanatory
selection.
The problem of explanatory selection comes from the fact that we do not deem
all causal information as explanatory relevant (Ylikoski 2001: 20; Hesslow 1983). We
can describe a vast array of events and processes leading to an event we are interested in
explaining just as Hempel points out. How then to select which of the factors of the
causal history to include in our explanation? What factors count as truly explanatory?
The attempt to describe the whole causal history would be practically impossible, as
Railton himself also acknowledges, but also if one were to know the whole causal story
leading to an event, that would still not count as a proper explanation. Thinking in terms
of the phrase “more causal knowledge the better for the explanation” might lead to
situations of pragmatic explanatory failures (Ylikoski 2001: 26) where one simply gives
too much information in answering a certain question. If I am interested in knowing
why John F. Kennedy was elected President in 1960, it is no use telling me the whole
history of the U.S presidential elections although the events related to the creation of the
election system surely belong to the causal processes leading to Kennedy’s election.
The problem of explanatory selection is common to all fields of study. A
historian who wants to write a traditional narrative of some event of interest and explain
the event faces the problem of choosing what parts to include into the narrative and
what to leave out (Little 2010: 14–15). As Daniel Little (2010: 30) puts it:

37
 “[A] crucial and unavoidable feature of narrative history is the fact of
selectivity. The narrative historian is forced to make choices and
selections at every stage: between ‘significant’ and ‘insignificant’,
between ‘sideshow’ and ‘main event’ and between levels of description.
(Is World War II better described at the level of generals and policy-
makers or infantrymen and factory workers?)
(Little 2010: 30.)

The contrastive approach offers a simple way of replying to Little’s question. If one is
interested in explaining some aspects of World War II, using contrastive questions helps
one define what factors are causally significant and what are not and what is the right
level of description since it depends on the contrast one is asking. In the Kennedy case
the information on the creation of the U.S presidential election system is irrelevant for
the question why John. F. Kennedy won the presidential election of 1960 instead of
Richard Nixon, but it might be relevant for other contrastive questions. If I were keen
on knowing why the U.S. has an election system where it is possible for Kennedy to
win the elections although he did not win in over 50 percent of the states, the
information on the creation of the U.S. presidential election system would be of high
interest, since it would explain why it is so that the U.S. has the electoral system it does
instead of some other system. The question concerning Kennedy’s victory presupposes
that there exist some electoral system in the United States, but explaining this
presupposition is giving answer to a different question than the question why Kennedy
won instead of Nixon.
The contrastive approach therefore implies that we do not need to be able to list
all the complex causal processes leading to an event in order to explain a certain
contrast. We only want to know those causal factors that account for the difference
stated in the contrast. What is the fact a that makes the difference between b and the
contrast b*? Contrastive explanatory questions are therefore means to nail down the
causal background4 of the explanation. By causal background I mean the causal
processes and events that one does not need to state or even know in order to explain
why b rather than b*. This is because in the contrastive approach the causal background
consists of the processes that b and b* share in common. For instance in the Kennedy
case the U.S electoral system is the same for the fact “Kennedy won” and its contrast
“Nixon won”. Therefore there is no need to state the development of the electoral

4
The causal background is often also called causal field following Mackie (1974: 34–35)

38
system in the explanation, since it does not account for the contrast of the explanation-
seeking question. In this case, the features of the electoral system and their development
belong to the causal background.

4.1.3!Negative!causes!and!the!principled!epistemological!objection!reconsidered!
The reader might remember the principled epistemic objection against the use of
counterfactuals in historiography presented in section 2.2.3. According to the objection,
we cannot know the things that did not happen, since we do not have a causal access to
them. The antagonists of counterfactuals consider this to mean that we cannot have any
knowledge of the antecedent or consequents of counterfactual claims since, by
definition, the antecedents and consequents of counterfactuals are things that did not
happen. Thus it is beyond the reach of empirical sciences to arrive at informed
conclusions of the truth-values of suggested counterfactuals.
I think that we are now in a position to answer the principled epistemic
objection. Counterfactuals that are conceptually linked to causal claims are
epistemically respectable, since causal relations that the causal claims refer to are
entities in the world that we have an epistemic access to: historiographical evidence can
be used to make inferences on the past causal claims and the related counterfactuals. If
the change from a to a*!brings about the change from b to b* in the variable B, there is a
invariant causal relation between a[a*] and b[b*]. There is nothing counter to facts in
causal invariant dependency relations themselves (Ylikoski 2011: 165) – they are
relations in this world, and one feature of them is that they support interventionist
counterfactuals. Therefore, interventionist counterfactuals can also be said to have truth
values – another questions is, of course, whether we have enough evidence to infer the
truth values of individual counterfactuals.
The above solution also brings clarity to the fact that sciences seem to be full of
sensible claims about things that did not happen, and to which we do not have a direct
epistemic access, such as negative causes, gaps, omissions, preventions and absences
(Ylikoski 2001: 37). With the help of the idea of causal claims with contrastive
explanans and explanandum it is easy to see why negative causes do not prove to be

39
insurmountable for scientific causal reasoning. Take for instance the following causal
claim and the related counterfactual scrutinized by Lebow & Gross Stein (1996)5:

(4.1) The fact that Kennedy did not show enough resolve prior to the Cuban
missile crisis caused Khrushchev to deploy the missiles.
(4.2) Had Kennedy shown more resolve prior to the Cuban missile crisis it
would have prevented Khrushchev from deploying missiles in Cuba.

What is evident in Lebow’s and Gross Stein’s study is that they do not concentrate on
the verbal expressions such as ‘did not show’ or ‘would have prevented’. In fact, they
formulate (4.2) without the term ‘prevention’:

(4.3) Had Kennedy displayed greater resolve prior to the crisis, Khrushchev
would not have sent missiles to Cuba (Lebow and Gross Stein
1996:124).

This makes it clear that our worries should not center on words or the exact
formulations of sentences, but we should rather focus on the contrastive pairs implied
by the causal claim. Thus, from the perspective of contrastive counterfactual theory,
Lebow and Gross Stein are interested in the relationship of the contrast of the purported
cause variable [c = Kennedy not showing resolve; c* = Kennedy showing resolve] to the
contrast of the effect variable [d = Khrushchev deploying missiles; d* = Khrushchev
not deploying missiles]. It is also important to note that in this case the fact c can be
formulated either as an absence [Kennedy did not show enough resolve] or as a positive
instance [Kennedy displaying greater resolve]6. On the basis of the new evidence on
Khrushchev’s aims and beliefs, Lebow and Gross Stein end up concluding that the
counterfactual (4.2) is false. The right counterfactual has the contrasts other way
around: it was precisely because Khrushchev belied the United States to be superior,
and therefore, to pose a serious threat to the Soviet Union under the Kennedy
5
Lebow and Gross Stein do not present the causal claim (4.1) themselves, their interest lies in the
counterfactual claim (4.2), but as I interpret them, they seem to provide evidence that works for
determining the truth of (4.1) as well as of (4.2).
6
The contrastive counterfactual theory also gives a simple way of representing omissions, negative
causes, absences and representations. We can think of a variable A taking the value a, for instance a = 0,
that represents an omission or prevention or any other type of negative cause. The variable itself “exists”,
it just does not have any positive value in the case of negative causes.

40
administration that he decided to send the missiles to Cuba (Lebow & Gross Stein 1996:
135).
Lebow and Gross Stein seem to have quite a strong epistemic grip of the
counterfactuals they assess. My claim is that this is so because in the end their argument
concerns the actual causal relations that took place during the Cuban Missile Crisis. The
counterfactuals can be employed in order to study these relations by using the evidence
there is on Kennedy’s, Khrushchev’s and their advisors’ actual aims, goals and
decisions. The antecedents and consequents of counterfactuals are therefore not out of
this world and it is not beyond the reach of our epistemic capabilities to evaluate the
truth of suggested counterfactuals.
So, what is to be replied to A.J.P Taylor (1969: 16–17) who worries on
counterfactuals echoing the strong epistemological thesis presented in chapter 2: “How
can we decide on something that did not happen? Heaven knows, we have difficulty
enough in deciding what did happen”? A simple answer is that if there is enough
evidence that gives us knowledge on the factual causal relations, we are also gaining
counterfactual knowledge. What is more, there are cases as the Lebow’s and Gross
Stein’s study presented here in which starting from counterfactual claims is a fruitful
way to approach the causal relations one is interested in. Contrary-to-facts is one way to
study the facts.

4.1.4!Objectivity!and!subjectivity!in!historiography!
One question of importance in historiography has been the issue of subjectivity and
objectivity of historiographical research. A source of confusion in this debate has been
the fact that the terms ‘subjectivity’ and ‘objectivity’ have not been very well defined
(Martin 1989: 85–86). It is, indeed, so that the debate on subjectivity and objectivity can
concentrate on many different, but in some cases interrelated topics such as the role of
values in research and the role of interpretation in historiography. My intention is not to
get deeply immersed in these themes but instead to make a few comments on the debate
based on the contrastive counterfactual approach.
In the context of historiographical explanations surprisingly many historians
have contented that historiography is inescapably subjective, whereas philosophers have
been quick to defend the idea that in principle historiography can be objective (Martin
1989: 85–86). It is worth separating three issues in this context:

41
 (S1) Subjectivity in choosing research questions: There is no other way of
deciding what to study and what to seek explanations for than the
researcher’s interests and values or the interests or values of the research
community or society as a whole.
(S2) Subjectivity in selecting explanatory factors: the selection of explanatory
facts and factors that one includes in one’s explanation depends on
subjective considerations such as researchers’ tastes.
(S3) Subjectivity in evaluating proposed explanations: The goodness of one’s
explanation is determined on subjective grounds. This also implies that it
is a matter of subjective taste to prefer one explanation to another.

When it comes to choosing one’s explanation-seeking questions, the counterfactual

contrastive model fits together with the first account of subjectivity. Subjective or
pragmatic or value considerations can be said to play a role in choosing the topic of
one’s study and in formulating one’s (contrastive) explanation seeking questions.
When it comes to assessing counterfactual claims as causal claims, the
contrastive counterfactual model takes a non-subjectivist stance towards (S2) and (S3).
First of all, the idea of explanatory selection presented earlier makes the rejection of
(S2) quite obvious: once the contrastive explanation seeking question has been fixed,
the selection of explanatory factors is an objective matter (Ylikoski 2001: 36; Northcott
2008: 80). We should “pick” or select those factors and facts in the explanation that
make the difference between the occurred fact and its contrast. This also means
objecting to (S3): the goodness of the proposed explanation depends on the question
whether it is so that the facts that one has selected in one’s explanation account for the
asked contrast. When there are two or more competing and mutually exclusive
explanations that seek to answer the same contrastive question, there is no reason to
base the choice between competing explanations on subjective considerations: the
advice is to seek evidence that would support or falsify the explanatory role of the
proposed explanatory facts.
E.H. Carr (1964: 96) has dismissed all counterfactual reasoning in
historiography as the “might-have-been” school of emotion. But, even if a
counterfactual claim was derived from emotional reactions or other value-laden
considerations, the origins of the claim should not be used to condemn it as unscientific
or unhistorical, as Carr (1964: 97) does. When it comes to counterfactual questions like

42
“what if the Russian revolution had not been?”, the reason for presenting this kind of
question might derive from the emotional responses of the victims of Bolshevism as
Carr (ibid.) suggests. Another matter then is whether this kind of question is to be
interpreted as a causal question that seeks to assess seriously the causal importance of
the Russian Revolution for some aspects of the events that followed the Revolution (for
more elucidation on this, see chapter 6). The contrastive counterfactual theory regards
this question as a proper causal question. The alternative is just what Carr (1964: 97)
claims: that counterfactual claims and questions are parts of parlour-games – a form of
“iffing” that people constantly engage in in their everyday lives for amusement or for
comfort without any real interest in the truth of the counterfactuals as causal claims. But
not all counterfactual reasoning is of this second type, as Carr believes. Counterfactual
reasoning goes hand in hand with causal reasoning, and in this context the justification
of the counterfactual claims should and can go beyond subjective iffing, even if the
counterfactual claims would stem from clearly subjective interests.
By denying the subjective claims (S2) and (S3) I am now making a principled
point. This means objecting to the ideas according to which there are conceptual or
principled reasons why explanatory selection and evaluation of explanations are
inherently subjective in historiography or in all fields of research in general. My
intention is not to dismiss in any way the problems one encounters in actual
historiographical research, since there might be empirical reasons why the selection of
explanatory factors and evaluation of competing explanations can in some cases be
extremely difficult. This might be due to the lack of evidence and unsettled disputes on
what some specific piece of evidence is evidence for. There are also important
epistemological and methodological questions I have not touched yet. How to come to
the conclusion that the proposed factors truly account for the difference? On what kind
of inferences and evidence one should base one’s evaluations of the goodness of one’s
explanations? These questions will be further discussed in the next chapter.
Raymond Martin (1989: 53–63) defends the objectivity of historiographical
explanations in a very similar way as I have done here. Martin’s argument is targeted
towards William Dray who has claimed that the impossibility of distinguishing causal
background conditions from explanatory causes makes explanatory selection a matter of
subjective taste. Martin tackles Dray’s idea by suggesting:

43
 “… ‘the cause’ is selected from among the contributory causes on the
basis of comparison between a situation in which the result to be
explained occurred and some other situation which I shall call ‘the
comparison-situation’. … ‘The cause’ then, like the conditions contributes
to the result, but unlike the conditions, also differentiates between the
situation in which the result occurred and the comparison-situation.”7
(Martin 1989: 59.)

The idea of using a comparison situation or a contrast in specifying the explanation-

seeking question allows Martin (1989: 62) to conclude: “it is always possible to
distinguish causes from conditions on factual grounds by addressing an appropriately
formulated explanatory question”. Dray’s idea that historians are forced to distinguish
explanatory causes from conditions on moral or other normative grounds is thus
undermined.
There might be many historiographical studies in which the selection of causes
has happened on moral or political grounds as Dray suggest to have happened in various
interpretations of the American Civil War (1962) or in the debate surrounding A.J.P
Taylor’s interpretation of Hitler’s role in the outbreak of the Second World War (1978).
Dray’s idea (1978: 150–151) is that Taylor and his critics engage in causal reasoning in
the way that lawyers do: they seek to attribute moral responsibility, blame and
appraise8. Attributing moral responsibility seems to require causal reasoning, but unlike
Dray, I see it highly problematic to think that attributions of moral responsibility should
guide the causal reasoning of a historian. Following the method of a historical
prosecutor leads quite easily to the situation where only causal relations relevant to
moral attributions are considered as the only causal relations that matter. For instance,
in Taylor’s case the question what the causal role that Hitler played in outbreak of the
Second World War, has become merely the question of whether or not Hitler intended
to cause the Second World War. The question of Hitler’s intentions is of high relevance
when assessing Hitler’s responsibility in the start of the Second World War, but there
are causal processes leading to the Second World War that Hitler participated in and in

7
Martin (1989: 159) calls this view “the consensus view” which he bases on ideas of philosopher’s such
as C.J. Ducasse, J. Feinberg, S. Gorowitz, H.L.A Hart, A. Honeré, M. Scriven, R. Shope, and M. White.
8
The connexion between causal explanations and attributions of moral responsibility in the context of
law and in history is extensively discussed by H.L.A Hart and Tony Honoré (1985). Hart’s and Honorés
central idea is that historians often think of causal explanation in the same way as lawyers do.

44
which Hitler’s causal role can be determined without knowing his intentions or
motives9.
Therefore, there is a need to distinguish the two roles of a historian: the
researcher aiming at causal explanations and the prosecutor aiming at moral
convictions. If a historian seriously wants to act as a prosecutor, she should make this
aim explicit, bring forth the causal relations of the case she is working on, and then pin
down the causal relations that she considers to be relevant for attributions of moral
responsibility and the reasons she has for considering just these causal relations relevant
for solving questions of moral responsibility. The worst thing that a historian could do
is to act as a bad historical judge who interprets the evidence and selects the facts that
she deems as explanatory on the basis of a priori considerations of who is guilty and
who is to be praised. This kind of method is bound to lead to serious subjective biases in
historiography. However, as I, as well as Martin, have aimed at showing this is not the
method that a historian needs to embrace.

4.2.!Facing!the!competition:!non6causalism!and!covering!laws!

4.2.1!Intentional!explanations!and!causal!explanations!!
William Dray (1963) has argued, following Collingwood (1994[1946]), that
Geisteswissenschaften, the sciences of the mind, do not involve causal explanation.
According to Dray, the sciences of the mind, such as historiography, focus on the
behavior of human beings in so far as it counts as action or as an expression of rational
thought (Dray 1963: 108 ; D’Oro 2011: 144). The rational explanation of action entails
constructing an argument – a practical syllogism – that shows that there is a rational
conceptual relation between the agent’s action and the agent’s beliefs and
desires/motives/goals. In rational explanation the relation between explanans and
explanandum is therefore not an empirical or causal connection, but a conceptual
relation that is established through reflection or understanding (Dray 1963: 108; D’Oro
2011: 144). Other philosophers who have developed a non-causalist position analogous
to Dray’s and Collingwood’s include, to mention a few, Peter Winch (1990[1958]) and

9
For instance, Hitler’s command to attack to Poland had the effect of starting the process of declarations
of war that cumulated into the Second World War. In this case, we do not need to know Hitler’s actual
intentions in order to say that this act of Hitler sparked off the Second World War. For, the assessment of
Hitler’s moral responsibility it might be of interest to know what Hitler was actually aiming at by giving
the order to attack to Poland.

45
Georg Henrik von Wright (1971). How does this position relate to the contrastive
counterfactual theory?
First of all, the contrastive counterfactual account is easily adapted to give a
causal interpretation of intentional explanations (Ylikoski 2001: 96–100). Intentional
explanations concern practical reasoning where an agent deliberates based on her
desires, motives and beliefs what action to take. The contrastive counterfactual theory
deems as causally explanatory all factors that make a difference stated in the
explanandum. If the explanandum consists of the question why the agent H did the
action c instead of c*, we quite often consider the agent’s beliefs and desires to be the
factors that account for this difference. Thus, in the case of practical reasoning, the
contrastive counterfactual account is in line with our everyday activity of explaining
human actions (Ylikoski 2001: 96). We explain our own action as well the actions of
others with the counterfactual framework all the time: if she had believed differently,
known this or had her motives been different, she would not have acted in the way she
did. Therefore, a historian who wants to know why Kennedy did not want to launch an
airstrike against the missile centers in Cuba during the Cuban Missile Crisis instead of
taking the advice of some of his advisors and do so, is causally explaining Kennedy’s
action when she answers the question by citing Kennedy’s beliefs and goals.
It is important to notice that this analysis is compatible with the idea that human
action is not the same as the movement of mindless blocks: human action can said to be
conceptually connected to ideas of rationality, deliberation and freedom as non-
causalists take it to be. However, the contrastive counterfactual account refutes the
thesis “reasons are not causes” as an either-or distinction10. Reasons can be taken as
causes if they account for the contrast in the explanandum. By claiming that reasons can
be causes, I am not saying that the reasons would stop being reasons in the context of
causal explanations. Reasons qua causes have similar kinds of counterfactual
dependency relations to action than all other types of causes have to their effects. The
causal reasoning with reasons works with the same interventionist idea: If there had
been an intervention on the agent’s reason r[r*], she would have done action c* instead
of c. This interventionist interpretation does not imply that the agent’s reasons should be
subjected to wicked external manipulations performed by other agents. The intervention

10
Donald Davidson (1963) has extensively argued for this position.

46
can also be the agent’s own change of mind. Thus, the intervention can come from
various sources. All that is required is that it makes r[r*] a switch for c[c*].
I simply cannot discuss all the rich arguments that the non-causalists presented
to support their position that causal explanation does not coincide with explaining (or
understanding) human action. In any case, it seems that the contrastive counterfactual
account is compatible with the main idea of the non-causalists: there can be conceptual
relations between the explanans and explanandum but that does not exclude the
possibility that there can also be causal relations between them. It can also be that in
order to verify the reasons for a certain action we need to have an idea what the action
was about and vice-versa, as von Wright (1971: ch. 3) suggests, but this does not mean
that there could not be counterfactual dependency relations between reasons qua causes
and action11.
In order to grasp why the non-causalists had reasons to argue for a distinct
method for explaining human conduct, one needs to point out that the theory of
explanation that they contrasted their views with, was not the one presented here, but
the covering law model. It was the covering law model that many non-causalists took to
provide the framework for causal explanation12. Since the non-causalists regarded the
covering law model as an implausible candidate for explaining human behavior for
various sorts of reasons, they had a good case to argue for a non-causal theory of
explaining human action. So, let us turn to the covering law in order to grasp the theory
of scientific explanation that the non-causalists were opposing in the spirit of anti-
positivism and German tradition of hermeneutics.

11
Not to give a wrong impression, there is a need to stress the following: there can be cases where the
conceptual link between the explanans and explanandum is too tight to provide the kind of interventionist
counterfactual dependence that is required in causal explanations. For instance, the agent’s intention of
greeting a friend is conceptually too close to be an explanatory cause to the greeting the agent gave. This
would be tantamount to saying that I greeted because I wanted to greet, which is not very illuminating.
(cf. von Wright 1971: 123–124). However, there is no conceptually dubious connection between the
intention of catching the friend’s attention and greeting the friend. Therefore the contrast [the belief that
greeting a friend is an apt way to catch the friend’s attention; the belief that greeting a friend is not needed
to catch the friend’s attention] can be used to causally explain [the greeting; not greeting] without any
problems. It is true that the content of the belief conceptually constitutes the act of raising the hand as an
intentional action: it was a greeting, in contrast to a compulsory movement and other plain physical
bodily movements. But since the changes in the belief also bear right kind of counterfactual relations to
the explanandum, the belief can in this case presented as an explanatory cause.
12
It might be a mistake to interpret Hempel as providing a model for causal explanation, since the
covering law model does not operate with or try to explicate the concept of ’cause’ at all.

47
4.2.2!Debunking!the!covering!law!model!
In 1942, Carl Hempel published the article “The Function of General Laws in History”.
It was the first publication on the model of scientific explanation that came to be known
as the Hempel-Oppenheim-model, deductive-nomological model or the covering law
model. In philosophy of historiography it launched a debate that flourished especially in
the 1950’s, 60’s and 70’s, but which still seems not to have ceased. Articles and books
were written pro (e.g. White 1965; Murphey 1973; Roberts 1996) and against (e.g.
Scriven 1959b: 457; Donaghan 1964; Danto 1965: 211) the applicability of the covering
law model in historiography.
The covering law model is meant to provide a universal account of scientific
explanation that would apply to all fields: various natural sciences, social sciences, and
as explicitly advertised by Hempel in “The Function of General Laws in History” to
historiography. The covering-law account states that an explanation of a singular fact
is a valid deductive argument where the explanandum is deduced from and shown to
follow from the initial conditions C1…Cn in accordance with the law statements L1…Ln
(Hempel 1942: 36; Hempel & Oppenheim 1998[1948]: 207–209). Since the explanatory
potential of suggested explanations relies on the existence of laws from which the
explanandum is deduced with the help of the initial conditions, all explanations that do
not explicitly state laws are seen as incomplete explanation sketches (Hempel 1942: 42).
For instance, the covering law account denounces the singular causal claim ‘The
extinction of the dinosaurs was caused by a strike of a huge meteorite’ as an elliptical
explanation (Hempel 1965: 415) since it does not explicitly rely on any laws.
The covering law model immediately seems to raise at least three worries to any
special science that deals with token explanations. First of all, it seems that the
meteorite strike explains the extinction of the dinosaurs without any reference to laws.
Secondly, one can also ask what kind of laws each special science should use. Laws of
physics? Are there biological laws? Does historiography have laws of its own or should
historians borrow their explanatory laws from social and behavioral sciences? Is the
generalization (4.1) ”whenever a nation s subjected to the challenge of the sea, has
excellent soil, is assisted by its neighbors, and has an excellent maritime situation, it
will rise to the great heights of success” a historiographical law? Is it really needed in

48
order to explain why the Netherlands rose to prosperity in the 17th century? 13 Thirdly,
one can also ask what laws are in the first place.
My purpose is not to dwell on these questions in a detailed manner. They are
introduced in order to show that the covering law account seems to have raised more
questions concerning the methodology of historiography than it has been able to solve14.
For instance, were a historian to adopt the covering law model as his view of
explanation, it would not give her much of advice on what to do next. The covering law
model is committed to the idea that for every causal explanation, be it a token or type
level explanation, there underlies a structure of causal nomic connections that gives the
causal explanation its explanatory import (Hempel 1965: 360–361). This idea does not,
however, give a direct insight on whether the historian should seek to justify her
singular causal explanations by searching for the covering laws that the singular causal
explanation supposedly refers to. As already mentioned, the historian would also be
puzzled because she would not be given any advice on where to look for these laws if
they are needed: in historiography, social sciences or perhaps, in physics? In the cases
where a historian compares competing explanations the advice would be to favor the
explanation that has a general law mentioned in it explicitly. The question still remains
on how to deal with cases where the competing explanations are equally elliptical?
This methodological unfruitfulness of the covering law model can be seen in
Raymond Martin’s (1989: Ch. 2) discussion on the covering law model in
historiography. Martin starts by presenting the debate on whether historians should try
to find covering laws in order to justify their singular causal explanations. He rejects
Morton White’s and Murray Murphey’s claims that covering laws are needed in order to
defend historiographical explanations. Despite of this, Martin (1989: 23) still claims that
the covering law model is the right account of what counts as an explanation.
On the other hand, Martin (1989: 25) thinks that it is sufficient for a historian to
be able to defend her favored explanation against competing explanations in order to
rationally endorse it. As an illustration of this idea he discusses the explanatory debate
among archeologists and historians on what caused the massive population loss of the

13
Morton White, who has introduced the generalization (4.1) (1965: 23), thinks that one should be able to
defend the generalization in order to defend one’s singular causal explanation of the rise of the
Netherlands. For more discussion on the topic, see Martin (1989: 19–23). Morris’s idea closely resembles
the idea that Woodward (2003: 164) calls the epistemological thesis: one needs to know the laws
underlying a causal claim if one wants to reliably tell whether the claim is true.
14
see Kosso (2011: 23–24) and MacDonald & MacDonald (2011: 133–136) for more examples of
questions that Hempel’s model has raised in historiography.

49
Classic Period Lowland Maya civilization in the 9th century. But he does not make use
of the covering law model anywhere in his analysis in order to make sense of the
competing explanations presented in the debate. As already mentioned in the section
4.3, in his discussion of weighted explanations and subjectivity in historiography,
Martin adopts the contrastive idea, which has nothing to do with the original covering
law idea. In Martin’s methodological discussions on explanation in historiography, the
covering law model turns out to be a dead letter.
The shortcomings of the covering law model in important methodological
questions can been regarded as a consequence of its committal to what Woodward
(2003: 159) calls “the hidden structure strategy”. Epistemically speaking, a proponent of
the hidden structure strategy claims the following: causal claims explain in virtue of
providing information about the existence of some underlying hidden structures that are
the regularities that make up the causal nomic structure of the world (Woodward 2003:
179). The problem is that even if we interpreted the causal structure to be the laws of
current physics, most people engaged in causal explanation are not familiar with these
laws. In spite of this, people, scientists and historians alike, evaluate causal claims and
condemn some explanations bad and good all the time. It is quite implausible to suggest
that people do these evaluations based on the causal structure of the world that is
epistemically inaccessible to them (Woodward 2003: 179–180). Since in sciences and
historiography explanations are constructed and evaluated with the help of
experimentation, observations and theoretical knowledge that is not quite often
formulated as sets of laws, it is no wonder that the covering law model offers little
insights on the questions how things are really explained and different explanations
assessed.
At the more general level, the standard philosophical counterexamples
introduced against the covering model might also be taken as sufficient in themselves to
debunk the covering law model. The various counterexamples prove that the covering
law account does not give the sufficient or, worse still, the necessary conditions for an
explanation15. The reason why I have not extensively presented the counterarguments,

15
Wesley Salmon’s (1971: 34) counterexample of a man, Jones, who takes birth control pills is one
example on how the covering law model fails to give sufficient conditions for explanations: it lets the
irrelevant fact that Jones is taking birth control pills together with the generalization “all males who take
birth control pills regularly fail to get pregnant” to explain why Jones does not become pregnant. Michael
Scriven (1959a and 1959b) has extensively argued that the model does not provide necessary conditions
for explanations: there are singular causal explanations that do not cite any laws or regularities. For
instance, we can explain why Jones developed paresis by referring to his untreated syphilis, although at

50
but, instead, taken up the methodological unfruitfulness of the covering law model, is
that my aim is to reach a more positive conclusion than just debunking the covering law
model. The theory of explanation to be applied in philosophy of historiography should
have merits of solving some methodological problems instead of giving no real
guidance or even creating more methodological controversies. Based on the discussions
of this chapter, there are reasons to believe that the contrastive counterfactual theory is
this kind of methodologically fruitful theory.
As already stated in the third chapter, the contrastive counterfactual theory is
comfortable with singular causal explanations. If a singular causal claim exhibits
“systematic pattern of counterfactual dependence” (Woodward 2003: 191), it can be
deemed as explanatory. Thus, the contrastive counterfactual model does not require that
explanations should be valid deductive arguments and that explanation and prediction
should be symmetrical. Nor does it claim that laws are essential to explanations and that
explanatory power of singular causal claims is based on references to laws. In other
words, the contrastive counterfactual account drops the features that characterize the
covering law model on the basis that they do not pick out what is essential to
explanatory import: invariance under possible interventions.
However, dropping the central ideas of the covering law model would not be a
big merit from a historian’s point of view without the methodological insights that
contrastive counterfactual account provides. In this chapter we have seen the insights
that result from applying the idea of contrastive explanandum and contrastive causal
claims to the problems of explanatory selection, subjectivity of explanations and
negative causes. In addition, the contrastive counterfactual account manages to solve the
principled epistemic objection presented by the antagonists of counterfactuals:
historiographical evidence can be used to support causal, and therefore, counterfactual
claims. The task of the next chapter is to shed more light on the question how this is
done in practice. The next two chapters will also show how the account deals with and
clarifies the important methodological issues of causal weighting and narrative structure
of explanation as well as provides a methodology for counterfactual inferences.

the moment medicine is not able to state under which antecedent conditions and together with what
generalization, the occurrence of a paresis is highly probable. In addition, only a small percentage of
those who have untreated syphilis develop paresis, but the probabilistic version of the covering law model
– the IS model – says that the laws and the antecedent conditions should make the explanandum highly
probable.

51
5!!!Epistemology!and!methodology!of!counterfactual!inferences!
!!!!!!!

!!!!!!!
One ought to be able to associate with any successful explanation a hypothetical or
counterfactual experiment that shows us that and how the manipulation of the factors
mentioned in the explanation would be a way of manipulating or altering the
phenomenon explained.
– James Woodward (2003: 11)

According to the theory of causal explanation presented in the earlier chapters, causal
claims imply counterfactuals. This is not merely a semantic point, but also an
epistemological one. If one has causal knowledge, one should also be able to engage in
counterfactual reasoning and answer what-if-things-had-been-different questions (what
if-questions for short). This means that the more what if-questions one can answer, the
more causal knowledge one has. What if-questions are not answered without evidence
however. Only with proper kind of evidence can one make reliable causal and
counterfactual inferences. Thus, it is crucial to have and idea of what kind of evidence
historians can use when answering what if-questions. How does the evidence that is
available to historians limit the certainty of their causal and counterfactual claims?
This chapter begins with the questions of evidence and counterfactual reasoning
(section 5.1). The second part of the chapter (section 5.2) explores to what extent the
contrastive counterfactual theory and the idea of interventionist counterfactuals can help
to solve some of the most pressing problems of counterfactual inference that historians
have presented. The point is to shed some light on the issue of what kind of
counterfactuals to use and what to dismiss. This scrutiny also identifies the reasons why
counterfactual inferences sometimes seem to be insurmountable for historians. In
addition, the scrutiny reveals the difference between counterfactual inferences on what
could have happened and inferences on what would have happened. The latter
inferences are central for causal reasoning, whereas the first ones are not.

5.1!Epistemology:!Counterfactual!reasoning!and!the!use!of!evidence!!!!
When discussing counterfactual and causal reasoning, many current protagonists of
counterfactuals talk of “counterfactual thought experiments” (e.g Tetlock & Belkin
1996; de May and Weber 2003; Lebow 2010: ch.2). When conducting an experiment
one applies the method of varying one variable and leaving the values of other variables

52
intact in order to see what kind of effect this change has on the outcome of interest.
When conducting a counterfactual thought experiment the varying is primarily done, as
the term ‘thought experiment” suggests in the mind of the historian1.
The terminology of “counterfactual thought experiments” is slightly
problematic, however, since the talk of “experiments” does not reveal that there is a
crucial difference between the evidential roles of real experiments and thought
experiments. Real experiments in the sciences are conducted by actually controlling for
other variables in order to find out the effects of certain variables. These experiments
generate evidence that can be used to support causal hypotheses and to discriminate
between competing hypotheses. Instead, counterfactual thought experiments do not
generate any new empirical evidence. Counterfactual thought experiments are
inferences in which one makes use of the readily existing evidence in order to support
one’s claims of what would have happened if this and this variable had been different in
a certain way. In other words, counterfactual thought experiments cannot be used as
independent evidence to support causal hypotheses. They are means to explicate and to
scrutinize the implications of causal hypotheses and they provide a means to check how
much support a certain causal claim has in the light of the evidence. This evidential
impotency is a crucial difference between actual experiments and counterfactual
thought experiments, and therefore, in order to avoid confusion I talk of counterfactual
inferences instead of counterfactual thought experiments.
So, what kind of evidence historians make use of when assessing causal claims
and the related counterfactual inferences? Historians are not, in general, in the position
to conduct controlled experiments, which in the light of the theory of causal explanation
defended here do seem to provide the optimal, although by no means perfect, way of
answering what if-questions. In what follows, I will sketch one method that is useful for
causal reasoning and answering what-if-questions in historiography. Due to lack of
space I cannot focus on statistical inferences and statistical methods, which in some
fields of history like in economic or population history are a valuable tool for making
causal and counterfactual inferences2. Nor can I discuss the exotic methods like the role
of computer simulations in causally explaining the past3. Instead I will discuss the

1
This does not mean that external aids like pen and paper or even computer simulations could not be
used to help out in these inferences.
2
On the use of statististical methods in historiography see e.g. Jarausch & Hardy (1991) or Feinstein &
Thomas (2002).
3
On the use of computer simulations in counterfactual inferences, see Cederman 1996.

53
method of process tracing in the context of qualitative case comparison studies and
single-case studies4.
Simply put, process tracing is about finding the mechanism(s) that produced the
outcome of interest (Steel 2008: ch.9). Mechanisms are usually considered to be
repeatable or frequently occurring combinations of entities and their organizations that
interact according to some invariant causal generalizations to produce the outcome of
interest (see e.g. Hedström & Ylikoski 2010 for a general overview on the definitions of
mechanisms presented in the literature). In some fields of historiography the idea of
mechanisms as repeatable is easier to grasp and to accept than in some others. For
instance, in economic history it is common to suppose that the same kinds of market
mechanisms have occurred in various contexts at different times to produce similar
kinds of outcomes. However, in other fields of history the mechanisms that historians
are interested in are considered to be local, highly context-dependent or as Stuart
Glennan (2010) puts it: ephemeral. Therefore, repeatability is not a property that
mechanisms studied in historiography necessarily have: a mechanism that a historian is
interested in can also be a one-time combination of entities and their organizations that
interacted according to some invariant causal generalizations5 to produce the outcome
of interest (ibid). Let’s discuss two examples to highlight the ways in which process
tracing works in comparative case studies and single case studies to provide evidence
for causal and counterfactual claims.
Katalin Miklóssy (2007) presents a comparative case study on Hungary’s and
Romania’s changing roles in the international politics during the Cold War. The
comparison begins from the year 1956 and ends in the late 1970’s, when the
consequences of the OSCE-meetings in Helsinki (1975) and Geneva (1973-1975) had
become apparent. Miklóssy asks why Romania managed to lose its good reputation in
the minds of the Western leaders during 1970’s, whereas Hungary managed to leave
behind the isolationist consequences of the uprising of 1956 and to become the West’s
favorite socialist country during the OSCE-process? In the 1960’s Romania was able to

4
The methods mentioned are not mutually exclusive; they can all be used to together in order to arrive at
conclusions on causal relations.
5
The term ’generalization’ does not only refer to patterns that repeat in various different context.
Invariant causal generalizations can also be context-specific: they describe how things work or are done
in general in a certain context. Therefore, also social rules of conduct that tell how people or certain
inviduals behave in some cultural or social contexts are also interpreted as invariant causal
generalizations. They can be used to support causal claims: the brothers-in-law in the Trobriand tribe give
presents to the husbands of their sisters, because brothers-in-law [instead of grandfathers or sisters-in-
law] are expected to give presents to the husbands of their sisters.

54
break the leash of Soviet control by conducting independent economic and trace
policies and it became the first “satellite country” to form official trade relations to the
west. During the 1970’s, however, both the Soviet Union and the western countries
turned their backs on Romania. Hungary, on the other hand, abandoned its radical
economic reforms in 1968 in the fear of another military intervention by the Soviet
Union. Despite of this drawback in the late 1960’s, Hungary managed to quietly
restructure its economy and and to form 55 new bilateral trade agreements with western
countries during the years 1976–1977.
Miklóssy’s method of finding the factors that account for this difference is to
compare Hungary’s and Romania’s foreign policies, and to point out the differences in
these policies that in the changing international environment lead to different kinds of
outcomes. Her conclusion is that Hungary was able to gain a good reputation in the west
by stressing the importance of human and minority rights issues in international arenas
in the 1970’s. The 1970’s was the period when the western countries also raised these
issues onto their political agenda. This new emphasis on human rights reflected
negatively to Romania’s reputation: Ceausescu’s Romania came to be known as the
brutal totalitarian regime suppressing the rights of its citizens in a Stalinist manner. The
loss of reputation was the cause why the western countries decided not to renew their
trade agreements with the country.
Miklóssy’s explanation for why Hungary managed to keep up good relations
both to Soviet Union and to the West, was its tacit way of operating in international
arenas. Officially, Hungary followed the lead of the Soviet Union, for instance, by
backing up the occupation of Czechoslovakia in 1968. This manner of conducting gave
Hungary enough space to quietly fulfill its own policy goals. Ceausescu’s loudness in
stressing the independence of Romania annoyed the Soviet leaders to the extent that
when the western countries turned their backs on Romania, the Soviet Union was not
there to help out.
Miklóssy’s conclusions give support to several counterfactual claims, although
she herself does not formulate her results in counterfactual terms. For instance, one can
conclude as follows: if Romania had managed to play by the rules set by the Western
countries and the Soviet Union in the same way that Hungary did, it is likely that the
country would not have ended up as an outcast in international politics. The factual
support for this counterfactual comes from the contrast between Hungary’s and
Romania’s conduct in international arenas and the different outcomes that the differing

55
ways of conducting led to. Hungary played by the rules set by the more powerful ones
and was successful in quietly transforming some of the rules to suit itself, whereas
Romania wanted to play by its own rules and was left in the position where it was in no
position to change to rules of those with the most power.
What makes Miklóssy’s comparative case study an example of comparative
process tracing? First of all, the study is motivated by the observation that Romania and
Hungary were both small socialist countries that despite the somewhat similar starting
points ended up occupying different roles in the field of international relations. The idea
is to explain the difference of outcome by comparing the mechanisms M1 and M2 that
lead from the similarity of the starting point A to the difference in the outcome B
[Hungary’s popularity in international arenas; Romania’s unpopularity]. In other words,
what is of interest in process tracing, is the causal relation between A and B, and the
causal and counterfactual inferences on the relation between A and B are supported by
the evidence resulting from exploring mechanism(s) Mn that link A and B together. The
assumption is that the causal relations within the mechanisms are easier to find or to
infer than it is to directly infer the holding or the nature of the causal relation between A
and B (Steel 2008: 187). Therefore, the mechanisms linking A and B are the means to
infer what is going on between A and B and thus the knowledge on the mechanisms can
be used to support causal claims on the relation between A and B and to explain B.
What is needed for the process tracing to work is6 identifying i) the entities of
the mechanisms = various countries in the field of international relations, most notably
Romania and Hungary, the Soviet Union and ii) their organization = the division of
power between small countries and superpowers iii) the invariant causal
generalizations that guided the interactions among the organized entities = the “rules”
according to which international relations operated in the 1960’s and 1970’s that
defined what small countries and superpowers were able to do7. Comparing the
mechanisms M1 and M2 where the two countries reacted in different ways to the “rules”
of the game supports causal and counterfactual inferences that explain the contrast in B
[Hungary’s popularity in international arenas; Romania’s unpopularity].
Careful process tracing on a single case can also give a historian enough
knowledge to do counterfactual inferences and to compare the actual case with the

6
See Steel 2008: ch. 9 for a detailed analysis on how process tracing works.
7
These rules did not stay the same during the years 1955–1980; were in constant change due to the
interactions between entities and it was precisely these changes that affected Hungary’s and Romania’s
positions in international relations.

56
inferred counterfactual case (and in this way to compare cases). I use Steel’s (2008:
189) discussion on Malinowski’s (1935) study on the wealth and influence of Trobriand
chiefs to illustrate this. Malinowski’s causal hypothesis was that the possession of many
wives was the cause of wealth and influence among Trobriand chiefs in the early 20th
century. Based on this causal hypothesis one can also make the following counterfactual
claim: if a certain chief had had more wives, he would have been richer. According to
Steel (2008: 189), Malinowski supports his hypothesis between the number of wives
and wealth by tracing the social mechanisms that link these two together.
The process tracing on the causes of the Wealth of the Trobriand chiefs reveals
i) actors (the entities of the mechanism) that are ii) organized in accordance to the social
roles attributed to them (e.g. chiefs, wives and brothers-in-law) and iii) invariant causal
generalizations that govern the interactions between the actors that are organized based
on their roles: the social rule is that brothers-in-law are supposed to give presents to the
husbands of their sisters – the gift is larger when the sister is married to a chief and the
chiefs usually have many wives. The knowledge provided by i), ii) and iii) supports
Malinowski’s causal claim. This example is taken from cultural anthropology, but
historians’ single-case studies also often rely on this kind of methodology of process
tracing: the aim is to explain why a certain thing happened in this way instead of some
other way by tracing the causal mechanisms that worked in a certain historical context
and accounted for the contrast of interest.
As the reader might have noticed, this section has contained a lot discussion of
evidence without clearly specifying what is this evidence that historians lean upon in
justifying their causal claims. Process tracing is one way to use evidence to support
causal and counterfactual claims, but what kind of evidence is needed in process
tracing? What kind of evidence is used in historiographical causal reasoning in general?
First of all, historians’ evidence consists of traces of the past. The traces can be
anything that has the potential to tell something of the time of interest8. For the purposes
of process tracing, a historian needs traces of the past that help her to infer i) what the
central actors and entities were ii) how they were organized iii) what kind invariant
causal generalizations there were to guide the interaction between the actors and other
entities. Ideally, knowledge of the mechanisms that were present at the time of interest

8
“Source critisism” is the method that historians use to check whether the source dates actually from the
time of interest and gives reliable information of the issue of interest.

57
helps historian to formulate and also to support various causal and counterfactual
claims.
Knowledge of the mechanisms that is needed in process tracing to support
causal claims does not merely stem from the traces of the past. A historian can also
apply her general theoretical knowledge of how societies, markets and humans behave
into the case one is studying in order to make causal inferences. This kind of general
knowledge that is applied into the cases that a historian is working on form the other
major source of evidence for a historian. General theoretical knowledge can be used to
supplement the knowledge inferred from the traces of the past or to interpret the traces
of the past in order to get a fuller picture of the causal relations and mechanisms of the
past. When a historian applies her general theoretical knowledge, she supposes that the
same mechanisms can work in variety of different contexts to produce similar
outcomes. For instance, the applications of rational choice theories to various historical
cases are an example of this kind of applying (see e.g. Bates et al. 1998).
Historians often have also various implicit theoretical ideas on invariant causal
generalizations and on how things work. They often also implicitly rely on these ideas
in reconstructing and inferring causal relations of the past9. Historians’ theoretical ideas
often stem from the views presented in the earlier history writing. They can also be
based on the theories of society and human behavior from various social and behavioral
sciences, as well as, the statistical and qualitative studies from these sciences. In
addition, historians quite often make use of their everyday beliefs about how do people
behave. When these ideas are not spelled out explicitly, the they are applied implicitly:
one makes causal claims on the past causal relations without specifying all the evidence
one uses to support these claims.
However, being serious about causal and counterfactual reasoning means that
one should be as explicit as possible about the evidence that one has in support of one’s
claims, since the “strength” of the evidence determines the reliability of the
counterfactual inferences. If one wants to do serious work on the past causal relations
one needs to spell out not only the evidence that stems from the time of interest, but also
where one’s general theoretical ideas of the working of societies and behavior of
humans come from – at least in cases where these theoretical ideas shape the causal

9
The prominent Finnish historian on international relations, Juhana Aunesluoma, (2010: 64–65) describes
the implicit use of theories in historiography as the method where “the theories are baked inside the
narratives”.

58
claims that one makes. One should not hide behind “historian’s imagination” that helps
one to make coherent interpretations of the past causal relations, but rather try to
document the evidence that one has in support (and against) certain causal claims. By
explicitly engaging in counterfactual reasoning and by representing the evidence in
favor of counterfactual inferences is an efficient way to try to see how much evidence
one has in support of one’s causal claims. Being explicit also makes it easier also for
other historians to intersubjectively to assess the quality of one’s work.
The fact that the justification of counterfactual inferences depends on the
evidence that there is to support the inference, reminds us of the epistemological doubts
of the antagonists of counterfactuals that I presented in the second chapter. The modest
epistemic objection says that there might be cases where the lack of evidence makes it
impossible to assess the truth of the proposed counterfactuals. The classic example of
Cleopatra’s nose serves as a nice example of this point. Is the counterfactual
“Cleopatra’s nose was the cause of the World War I” (see: Fearon 1996) to be deemed
as true or untrue10? My answer is that based on the evidence that we have at the
moment, we do not know. We simply do not have enough evidence on the causal
mechanisms from the time of Cleopatra to the beginning of the First World War. Some
of us might have some metaphysical ideas of the nature of history as a process where
everything affects everything, which would mean that Cleopatra’s nose would be the
cause of everything that has happened after the time of Cleopatra. On the other hand,
some of us might have intuitions that perhaps everything does not affect everything.
Cleopatra’s nose might have affected the behavior Mark Antony and Julius Caesar and
the wars they started, but perhaps these wars did not make a big difference in the long
run. In any case, the truth of the counterfactual “Cleopatra’s nose was the cause of the
World War I” is an empirical question. This is why we should not base our assessment
of this counterfactual merely on our metaphysical intuitions, although we would lack
sufficient empirical resources to reach any firm conclusions on the matter.
An antagonist of counterfactuals is surely right in pointing out that the lack of
evidence might prevent us from deciding on the truth-value of some counterfactuals.
For some counterfactual inferences, on the other hand, the situation might not be that
despairing. Absolute certainty cannot be expected from counterfactual inferences, since
10
Blaise Pascal introduced the well-known example of Cleopatra’s nose in his Pensées with the
counterfactual: "Cleopatra's nose, had it been shorter, the whole face of the world would have been
changed” (Pensées II, 162).

59
absolute certainty is out of the reach of any empirical science. So there is always some
degree of uncertainty in every counterfactual inference. By careful use of evidence, one
can try to assess just how reliable one’s counterfactual inferences are. This is the only
way to avoid the pitfalls of presenting one’s counterfactuals more strongly supported
than they are and to separate fanciful “iffing” from serious empirical work.

5.2!Interventionist!methodology!for!counterfactual!inferences!
One source of problems for a historian working with causal explanation and
counterfactual claims is the fact that evidence can be scarce, shaky and non-decisive.
Other bundle of problems descends from the difficulties of deciding on how to go about
with counterfactual inferences. What kind of things should a historian focus on, and
what kind of things to neglect when one tries to determine the truth of her
counterfactuals with the help of evidence? This section deals with this question by
looking more closely to the properties of interventionist counterfactuals.

5.2.1!Features!of!interventions!
Raymond Aron (1961: 160–161) provides a general illustration of the general idea of
how one should proceed in counterfactual inferences when one wants to determine the
causes of events:

STEP 1. Analysis of the phenomenon-effect.

STEP 2. Discrimination of the antecedents and isolation of one antecedent, the
efficiency of which it is desired to measure.
STEP 3. Construction of unreal developments.
STEP 4. Comparison of mental images and the actual events.

In the light of the contrastive counterfactual theory Aron’s idea could be interpreted as
follows:

STEP 1.* Analysis of the phenomenon-effect = define the contrastive explanandum

b[b*].
STEP 2.* Discrimination of the antecedents and isolation of one antecedent, the
efficiency of which it is desired to measure = If the fact a[a*] is
hypothesized to be the explanatory cause of b[b*], imagine the
intervention that changes a to a* and leave the other causal routes that
lead to the variable B intact.

60
 It is important to note that these conditions describe ideal interventions and ideal
interventionist counterfactuals. It might be that no intervention carried out in scientific
experiments or no counterfactual inference in historiography is able to fulfill all these
conditions. This does not mean that interventions that fall somewhat short of fulfilling
these conditions would not give any causal knowledge. The conditions of ideal
intervention can be seen as an regulative ideal (Woodward 2003: 114): they tell what
we should aim at in our investigation, since they tell what must be true in the
relationship between A and B if A causes B.
In what follows I will go through the conditions one by one in order to see to
what extent historians’ counterfactuals fulfill these conditions. I will analyze why
certain counterfactuals fail to fulfill these conditions and explore the consequences of
these failures. Even if a counterfactual fails to fulfill these conditions, it does not mean
that the counterfactual should be dismissed out of hand. The analysis in section 5.2.2
reveals that, in fact, there is no algorithm that would help us easily to discern
counterfactuals with “dubious” antecedents from counterfactuals with more respectable
antecedents. Sections 5.2.3 and 5.2.4 argue that sometimes historians’ object of study –
the causal relations in the past – makes it practically impossible for a historian to
conduct counterfactual inferences that fulfill the conditions 2 and 3. It is hard to make
inferences that fulfill the conditions 2 and 3 if the causal system under study resists to
work in accordance with these conditions: complex systems quite often require complex
inferences, and non-modular systems often lead to non-modular inferences.
Section 5.2.5 deals with the condition 4, and argues why it is always sensible to
strive to make counterfactual inferences that fulfill this condition if the aim is to make
causal inferences on the relation between A and B. Historians quite often make
counterfactual inferences that do not fulfill this condition, and it is argued that that these
inferences do not contribute to causal reasoning in any straightforward way. In the end,
the analyses of sections 5.2.2–5.2.5 sketch a picture of what historians causal reasoning
is like, and compare it to how it should be like. With what kind of inferences is it
reasonable for a historian to explore causal and explanatory relations?

62
5.2.2!Intervention(does!not!change!the!explanandum!variable!directly:!the!problem!
of!selecting!sensible!causes!
The need for a condition 1 (the intervention I does not change the variable B directly) is
quite easy to understand: if we want to test whether A is the cause of B, it makes no
sense that the intervention would change B directly. A good example of an intervention
that does not fulfill the condition 1 is a randomized drug experiment where the subjects
in the treatment and control groups are let to know whether they are treated with the
drug or with a placebo and where this knowledge of the intervention affects their
probability of recovery from their condition. Are there interventions and counterfactual
inferences that do not fulfill this condition in historiography? This question turns out to
be a tricky one, since it is not obviously clear how the condition 1 should be interpreted.
One group of counterfactuals that do not seem to fulfill this condition are those
where changing the antecedent from a to a* would “annihilate” the contrast b[b*]
altogether. Two following two counterfactuals seem to work this way at first sight:

(5.1) If Romania had been a superpower like the Soviet Union and the United
States [in contrast to being the small country it was], Romania would have
been able to resign from the Warsaw pact in the 1970’s [instead it stayed a
member of the alliance although it tried not to take part in its operation].
(5.2) If Napoleon had had Maxim machine guns in the battle of Waterloo [which he
did not have], he would have won [instead of losing]11.

How to treat these counterfactuals? According to one line of thinking the counterfactual
inference would go like this: let’s imagine that we do an intervention where we change
the status of Romania from a small East block country to a superpower. This
intervention means changing the history of the country significantly in such a way that
the country would not have become a member of pact dictated by the Soviet lead in the
first place. Thus, the contrast implied in the consequent of this counterfactual would be
have been left unrealized and the whole question of Romania’s status in the Warsaw
pact become pointless if the antecedent had been realized. The antecedent and the
consequent do not seem to fit together: if the antecedent realizes, there are reasons to
believe that the consequent cannot realize. The same applies in the case of the
11
This example is adapted from Fearon (1996: 40). Maxim machine gun was invented in 1885 by Sir
Hiram Maxim and was used extensively in the First World War.

63
counterfactual (5.2). If the situation had actually been so that Maxim machine guns
existed in the beginning of the 19th century, the situation in international politics would
most likely been altered in a way that the Battle of Waterloo would not have been
fought.
Many protagonists of counterfactuals who have dealt with similar kinds of tricky
counterfactuals, have rejected counterfactuals resembling (5.1) and (5.2) by arguing
along the lines presented above (e.g Tetlock & Belkin 1996, Fearon 1996: 57). This
type of reasoning seems very plausible, but it is not the only possible way to go about
interpreting counterfactuals (5.1) and (5.2). The condition 4 set for ideal interventions
states that interventionist counterfactuals do “miracles” in the sense that they break all
the links between the causes of the antecedent (the variables U in the picture 5.1) and
the antecedent A. The point is to isolate the relation between A and B so that this
relation, and this relation alone, can by studied. We want to know what kind of effect
A, and A alone, has on B is and for this purpose we want to manipulate A, not the
causal history of A, with the help of an exogenous intervention.
Therefore, according to the condition 4, when we are thinking of possible
interventions we should not concern ourselves greatly with the questions on how the
interventions could actually have realized or could they have been realized. I will argue
further why it is better to interpret counterfactuals along the lines of the condition 4 in
the section 5.2.5. But for now, however, in order to pursue further with the topics of this
section, let’s explore how the counterfactuals (5.1) and (5.2) are be interpreted if we
follow the condition 4.
In order to imagine an intervention in the international status of Romania, we do
not need to rewrite and to imagine to whole history of the 20th century and to create an
alternative history where Romania becomes a superpower by the 1970’s and to
speculate whether Romania would have become a member of the Warsaw pact in this
process or not. What we, according to the condition 4, need to do is to postulate that
Romania would have suddenly become a superpower with the central properties of
superpowers12 in one night, let’s say, in 1975, and ask whether this sudden change of
status would have helped Romania to retreat from the pact. Based on what we know of
the dynamics of international relations in the 1970’s, it seems to plausible to suppose
that a superpower could have resigned from pacts like the Warsaw pact or NATO. So, if

12
E.g. economic, military, technological, cultural leadership and due to this factors the power to work in
the field of international relations relatively independently.

64
we interpret the counterfactual (5.1) in this manner, it is not pointless, senseless or
absurd – it is, given the things we know, most likely true. This result applies also to the
counterfactual (5.2). If Maxim machine guns had suddenly fallen from the sky and
Napoleon’s troops had suddenly known how to use them, they would have won the
battle of Waterloo.
If we belief that counterfactuals (5.1) and (5.2) reflect invariant dependency
relations, it means that the fact that “Romania was not a superpower in the 1970’s”
should be regarded as a cause for the fact that “Romania did not break free from the
Warsaw pact”. This sentence might be considered slightly odd at first, but when we
realize that “Romania was not a superpower in the 1970’s” means in this context the
same as the sentence “Romania was a small country”, the counterfactual inference
becomes more comprehensible: the fact that Romania was a small country instead of
being a superpower explains why Romania was not able to break free from the Warsaw
pact.
The fact that “Napoleon did not have Maxim machine guns” is also to be
considered a cause for his defeat at Waterloo. There is a right kind of invariance relation
between the intervention on the variable of Napoleon’s field weaponry [Maxim machine
guns; no Maxim machine guns] and the outcome of the battle [victory: defeat]. Despite
there is a right kind of counterfactual dependency relation between the variables, we do
not find it that sensible to say that the lack of machine caused the defeat. Why is this?
Why does the fact that Romania was a small country strike as a more sensible cause
than the fact that Napoleon did not have Maxim machine guns?
There are several other examples that resemble the Napoleon case. James
Woodward (2006: 24–25) discusses the following example: the counterfactual (5.3) “if I
was struck by a huge meteorite, I would not sit here writing this thesis” is apparently
true. Therefore, the fact that I am not struck by a huge meteorite explains why I am
sitting here writing this thesis. To further expand Woodward’s example the fact that the
Earth has not been struck by a huge meteorite seems to have been a cause to everything
that has happened within the past 65 million years or so.
The counterfactuals (5.2) and (5.3) point out an interesting problem in the
contrastive counterfactual theory of causal explanation. The above-mentioned
counterfactuals present invariant dependency relations, and therefore according to the
contrastive counterfactual theory we should consider them as causes. On the other hand,
we are quite unwilling to consider the lack of certain type of machine guns and not

65
being struck be a meteorite as causes. How come? The reason is not that these causes
are negative causes that cite omissions. As we saw in the previous chapter omissions
can reasonably be considered as causes in some cases.
The problem has not been left unnoticed among the supporters of the contrastive
counterfactual theory. Woodward (2003: 90) answers with the following remark:
“Causal judgments reflect both objective patterns of counterfactual dependence and
which possibilities are taken seriously.” In other words, we consider some variables and
their values as sensible or plausible causes, whereas other variables and their values we
reject as implausible and these considerations enter into our judgments on which causes
we are willing to take seriously and to present as causes13.
So, what informs our considerations on which causes are to be taken seriously
and which not? Why are not Maxim machine guns and meteorites sensible causes
although we belief that some interventions on these variables have the right sort of
counterfactual dependence the effect variables? Why is the fact that Romania was a
small country a sensible cause variable? There seems not to be clear criteria for
deciding which causes to count as sensible and which not. There can be some heuristic
guidelines, however. The variable of international influence that can take the rough
values [superpower; small country] seems to be a sensible cause variable since there is
evidence that

i) This cause variable existed in 1970’s and it took different values = countries
were labeled as superpowers, small countries, middle-sized countries etc.
based on their ability to affect the direction of international politics
ii) The field of international relations was organized so that the cause variable
worked (and still works) as a difference-maker in international politics: there
were mechanisms (the way the relations between countries were organized
and the rules that governed these relations) that ensured that possible
interventions on or changes in the cause variable would have made a
difference to certain effect variables.

13
This is not a problem that raises merely in the context of historiographical counterfactuals. This is a
general problem of the interventionist theory (Woodward 2003: 89), and something like the notion of
”serious possibility” is needed to explain why not all counterfactuals that correctly reflect the invariant
dependency relations, are considered as explanatory.

66
Based on this example, we can generalize the following criteria that seem to affect our
considerations of sensibility:

(C1) the existence criterion: the variable needs exist in the case under study or if
the variable exists, the suggested value of the variable needs to have existed.
(C2) the mechanistic criterion: there needs to be a mechanism or mechanisms that
ensure that (in principle at least) the variable works as a difference maker. In
other words we need to have an idea how the variable would work as a
difference maker in a certain context14.

So why are Maxim machine guns not a plausible cause for the Napoleon’s defeat? The
answer is simple: Maxim machine guns did not exist at the time of Waterloo: the
variable concerning Napoleon’s field weaponry cannot be given the value of maxim
machine guns. The existence criterion is not fulfilled. The meteorite example at first
sight seems to fulfill both the criteria (C1) and (C2): meteorites surely exist and we have
an idea how the falling meteorites can cause the damage they do: if I was struck by a
meteorite I would most likely die, since the mere speed of the collision would break
vital parts of my body. The problem with the contrastive explanans [I am struck with a
meteorite; I am not struck with a meteorite] is that it doesn’t seem to “hit” the
contrastive explanandum [I am writing my thesis; I am not writing my thesis] in a right
way. First of all, the contrastive explanans explains several other contrasts as well: [why
my lungs are working; why they are not working] or [why I am alive; why I am not
dead]. However, the explanans seems to fair rather poorly in explaining these contrasts
as well. What is lacking is the context that links the explanans and the explanandum
together in a right way. The criterion (C2) is not fulfilled after all.
Let us suppose that I have read in the newspaper this morning that there will be
huge meteorite showers in Helsinki this afternoon. Before reading the news on weather,
I was determined to have a day off and not to write my thesis. Reading the news,
however, made me change my mind. I thought that perhaps it is better to write and
finish my thesis this morning before the meteorite storm hits me or the room I use to
write. In this kind of scenario the explanans [I am struck with a meteorite; I am not

14
This criterion resembles Petri Ylikoski’s (2001: 41) mechanistic requirement: “a explains f [c] if and
only if (1) a belongs to a causal process leading to f, and (2) there is a causal mechanism m that ensures
that f occurs instead of c because of a”. The mechanistic condition also ensures that there is a chance to
do process tracing to see whether the suggested cause and the explanans are connected or not and if they
are in what ways.

67
struck with a meteorite] seems to be one plausible explanation the contrast [I am writing
my thesis; I am not writing my thesis]. This means that our background knowledge of
the context defines our considerations what to take as plausible causes and what not.
The knowledge on the context defines our beliefs whether there was a mechanism that
ensured that the suggested variable or its values worked as a difference maker in a right
way.
This result should not be of great concern to historians, however. It is true that
we select from the vast range invariant counterfactual dependency relations those
relations that we belief on the basis of our evidence to fulfill also the mechanistic
criterion. This means that historians should keep on doing what the have been doing:
they should try to analyze their evidence, the sources of the past and theories on how
things work, in order to create hypotheses about the causes that in the context under
study could have brought about certain outcomes of interest. Careful study of the
historical context of interest helps to postulate causes that fulfill the criterion (5.2)
without any further ado. If the evidence does not imply anything of the possibility of
Maxim machine guns or meteorite strikes in the context of the Battle of the Waterloo,
there is no point to hypothesize about these variables as causes for the defeat at
Waterloo.
Philosophers who might be interested in finding out clear-cut criteria to separate
sensible causes from the non-sensible ones, might not be too pleased with the analysis I
have given15. In any case, the above-mentioned criteria are at the moment the deepest
level of analysis I can go when trying to make sense why certain causes are not
generally considered as causes, and I believe that in the context of historiography they
suffice. The main point of this section should be clear, however: our causal judgments
on sensible causes do not entirely follow the objective invariant dependency relations
reflected by counterfactuals interpreted along the lines of conditions 1–4. Secondly, the
framework of contrastive counterfactual theory does not collapse entirely in the face of
this problem. Instead it helps us to get to the point where we can see in a brighter light
where the problem lies: it is not so, that counterfactuals with problematic antecedents
should be considered meaningless, as it often happens when we try to rewrite the
historical context so that the realization of the antecedent becomes plausible. Instead,
many counterfactuals with problematic antecedents can be said to have truth-values if

15
Also some historians and social scientists express the need to have this kind of criteria see. e.g Fearon
1996: 40–41 ; Tetlock & Belkin 1996: 23–24

68
 The methodology of intervening on just one variable and keeping other things
equal was not possible for our poor imaginary anthropologist or historian because
intervention by the British colonial rule proved to be a non-ideal one. Is the
methodology possible in historiography in other cases, however? Can there be
counterfactual inferences that neatly fulfill the condition 2? These questions cannot be
given a definite answer, since it depends on the nature of causal structure under
investigation what kind of interventions are possible. So the right kind of question is
what kinds of causal systems or networks of causal connections historians are interested
in. It seems that in many cases historians are interested in systems where it is hard to
intervene on some causal connections without affecting the other causal relations.
Richard Ned Lebow (2010: 56–57) puts the point as follows: “History is like a spring
mattress. If one spring is cut, or simply subjected to extra pressure, the others will to
varying degrees shift their location and tension”.
How should we interpret Lebow’s comment? At this point it is important to
distinguish between two properties that systems can have that both reflect on our ability
to do counterfactual inferences on these systems: complexity and modularity. Failures in
fulfilling the condition 2 are failures because they jeopardize the modularity of the
intervened causal system: if a system is modular (or our representation of the system is
modular) it is possible, at least, in principle to intervene on its variables without directly
changing the values of the other variables that affect the explanans variable (including
the middlemen variables that are between the intervened variable and the explanans
variable). Simply put, modularity concerns “the separability of causal contributions to
an overall effect” (Mitchell 2008: 699).
The aim of the interventionist strategy is to try to guarantee system’s modularity
and to salvage the separability of causal contributions of different variables whenever
the causal system under study allows it i.e when the system actually is sufficiently
modular. Therefore, the values of the other variables that are not between the explanans
and explanandum variables and that affect the explanans variable should be kept fixed
when the explanans variable is intervened on (Woodward 2003: 59). In addition, as the
condition 2 says, the middlemen variables should not be changed by no means other
than changing A: interventions should be so that they guarantee the possibility to
separate the causal contribution of the intervened variable A from the causal
contributions of the middlemen variables that the variables have independently of A
(more on modularity see e.g. Woodward 2003: 48–49, 327 -342).

71
 Complexity of a causal system depends on two dimensions: i) the number of
middlemen variables between A and B, and ii) the number and nature of the causal
relations between the middlemen variables. Complexity of a system is a matter of
degree that is influenced by the both of these dimensions. For instance, when there are
causal loops between the middlemen variables the system of causal variables is more
complex than in the case where the relations are one-way. Causal loops are interactions
where the value taken by one variable F1 causally influences the value of F2, which then
affects the value of F1 and so on. Complex systems are, therefore, bundles of causal
relations and variables that react to interventions in the following way: an ideal
intervention on one variable A affects the values of several “middlemen variables” F1 …
Fn, and these changes in values might also change the nature of the causal interactions
between the variables.
Modularity and complexity can be connected in two ways in relation to an
intervention I on A in regards to B: there can be modular complex systems (picture 5.5)
and non-modular complex systems (picture 5.6). Both types of systems provide
challenges to historians’ counterfactual reasoning. In modular complex systems an
intervention on A that fulfills the condition 2 is possible, but the intervention on A
affects so many middlemen variables that it is hard make out what will be or would
have been the effect on B16. Inferences on non-modular complex systems are hard not
merely because of complexity: they are further complicated by the fact that we cannot
intervene on A without directly intervening on some of the middleman variables. This
makes it impossible to distinguish the causal effect of A on B from the causal effects of
the middleman variables.

16
This applies to cases like historiography where we are in no position to actually observe the
intervention and the subsequent values of B. The change in B caused by the intervention is not observed
but inferred on the basis of our evidence.

72
5.2.4!Intervention!does!not!correlate!with!other!causes!of!the!explanandum!variable:!
the!problem!of!the!unknown!common!causes!
The condition 3 says that the intervention on A in regards to B cannot be the cause or
even to correlate with some other variable C that is a cause of B. This condition is also
based on the need to isolate the causal relation of interest in order to see whether A
causes B or not: we are not interested of the overall effect of A and C on B, but only on
the purported effect of A to B. For instance, in randomized experiments one needs
randomization to exclude the other factors C that are correlated with the intervention
and might affect B independently of the causal path between I, A, F and B. For instance,
the process that shuffles people into control and test groups in drug trials should not be
so that it places the persons with a stronger immune resistance into the other group and
the persons with a weaker immune resistance into the other (Woodward 2003: 97). This
kind of bias in selection of the test and control groups would lead to a result where the
experiment does not test what it is designed to test: the effect of the drug.
Both in quantitative and qualitative comparative case studies one often
encounters the problem whether there are unknown factors Zn that are a common causes
of X and Y and therefore account for the observed correlation between X and Y. Only
when one is able to rule out common causes for instance with statistical means or
process tracing, one can infer that there is a causal relation between X and Y. An
intervention on X does not fulfill the condition 3 if the intervention also changes the
unknown variable Z that causes systematic changes both in X and Y – based on these on
systematic changes between X and Y is possible to falsely conclude that there is a
causal relation between X and Y.
Process tracing that seeks to find out the mechanisms between X and Y is not
always to pathway to avoid the problem of unknown common causes that are affected
by interventions: for instance in his study on the wealth of the Trobriand chiefs,
Malinowski is not able to rule out, for instance, the possibility that social status (S) is
the common cause that affects both the number of wives (N) and wealth of the chiefs
(W). Malinowski is only able to infer that there is some kind of connection between (N)
and (W) (Steel 2008: 191–192). Picture 5.9. shows the situation where an intervention
affects both (N) and (S). If we are unaware of the variable (S) and of the fact that the
intervention on (N) also affects (S), it is in this case possible to conclude that (N)
causally affects (W) although it does not.

76
 It is a striking feature of the kinds of counterfactuals that are relevant to causal
and explanatory claims that although we require that they be true when their
antecedents are realized by interventions, any more detailed specification of the
way their antecedents are realized seems inappropriate and unnecessary.

Historians, however, seem not to agree with Woodward. Historians are

immensely interested in questions on under what conditions the antecedents of
counterfactuals could have been realized. Historians often present questions of the form
“could this and this had been different” instead of asking “would this had been different
if this and this had been different”. Could-questions and could-inferences seem to be as
common as would-questions. This is at least the conclusion one must draw when
reading through the two of the most important edited volumes that propagate the
benefits of counterfactual thinking for historians – Counterfactual thought experiments
in world politics and Unmaking the West: “what if” scenarios that rewrite world
history. In practice, almost all of papers in these volumes support the conclusion that
Julian Reiss has drawn: “When implementing a counterfactual antecedent, the historian
thus asks what conditions would have to be present in order for the antecedent to follow
from these conditions, and whether these conditions were likely” (Reiss 2009: 719).
The editors of the volume Counterfactual thought experiments in world politics
Philip Tetlock and Aaron Belkin (1996: 7 & 18) compress their idea that historians
should think about the conditions that are needed for the realization of the antecedent
into a minimal rewrite rule: one should only present antecedents that require altering as
few “well-established” historical facts as possible. The rule can be interpreted so that
one should only use antecedents that “do not require us to unwind the past and rewrite
long stretches of history” (Tetlock & Belkin 1996: 23). This interpretation reveals
historians’ way of thinking about counterfactuals in a nice way: Tetlock and Belkin take
it as a given that historians “rewrite” history in order to see on what conditions and how
easily the antecedent could have been realized. For Tetlock and Belkin it is perfectly
acceptable that a historian rewrites the past in order to create story where the antecedent
of the causal claim is imagined to have a causal history that guarantees its realization.
Rewriting history means that when a historian imagines an intervention on the
variable A and imagines the intervention from a to a* as shown in the picture 5.1. he
turns to the causes U of A in order to see whether they would allow the intervention that
changes a to a* to actually take place. If it seems that the causes U would not allow the
intervention to take place (in the light of U the change from a to a* seems implausible),

78
the historian considers what changes there needs to be made in U in order to bring about
the intervention. Then the historian might even continue moving backwards in the
causal chain and look at the causes V of U in order to consider how easily the changes
in U could have taken place etc. Thus, historians’ counterfactual thinking sometimes
works by looking back in the causal chain for the causes of the supposed cause a. This
causal knowledge is, however, supplemented with extra considerations that are
informed by beliefs and assessments on how easy or difficult it could have been to
change the causes U of a. Together with causal knowledge, the assessments of
“easiness” of changes are used to estimate whether a could actually have been changed
to a* in the way the counterfactual says. Many of historians’ counterfactuals are could-
counterfactuals on possibilities: what could actually have happened? The interesting
property of them is that they are backward-looking in the way described above.
The contrastive counterfactual theory (the condition 4) states that there is no
need to use backward-looking counterfactuals in our causal inferences. The
considerations on whether the change from a to a* was actually possible21 or not should
not play a role in choosing explanatory counterfactuals. According to this theory a
historian should only be concerned about what happens after the intervention on a
certain variable in a causal system (the things that happen might show quite a lot of
complexity as we saw in section 5.2.3.). The probability of the actual occurring of an
intervention does not determine whether the intervention presents an interesting
counterfactual claim from the point of view of causal explanation.
Let’s illustrate the matter with an example before turning to the reasons why it is
always sensible to follow the condition 4 when making causal inferences. Yuen Foong
Khong (1996) asks whether it would have made a difference to the beginning of the
Second World War if Britain’s Prime Minister Neville Chamberlain had acted more
resolutely in the face of Hitler’s Sudetenland demands. This question concerns the
causes of the Second World War: did Neville Chamberlain’s actions in Munich play a
role in the breaking out of the war, and if they did, what kind of role it was? According
to the contrastive counterfactual theory, what a historian should do in answering this
question is to try to infer what would have resulted if Chamberlain had acted differently
in Munich. Inferring what would have resulted does not require that the historian needs
to think whether Chamberlain could actually have acted differently. Instead, he needs to

21
The actual possibility of an event is assessed by means of could-inferences, and the assessments are at
least in part based on historians’ estimations what could easily have happened and what not.

79
find evidence that tells how the Germans would have reacted if Chamberlain had acted
differently and what kind of changes in the field of international politics these reactions
would have caused. For instance, documents revealing the strategies that the Germans
were visioning in response to the acts of Britain and France would provide this kind of
evidence.
Yean Foong Khong presents the explanatory question on the causes of the
Second World War in the beginning of his essay and uses less than four pages to review
the evidence that helps him to answer this question. The remaining 18 pages he uses to
assess whether it could have been possible for Chamberlain to act differently, and if
Chamberlain could not have acted differently, could there have been another person as
Prime Minister who could have abandoned the appeasement policy in Munich or before
that22. Why is Khong like the other authors of the volume so keen on pondering the
conditions of realization of their counterfactual antecedents23? Does the pondering
contribute to the task of causal explanation that the writers take up? If it does, are
historian’s practice and the theory of causal explanation propagated in this thesis in a
clash? Does the clash mean that the theory of causal explanation is utterly wrong in one
of its main presuppositions? Or alternatively, are the historians who believe that they
are in the quest of finding the causes of events by pondering the plausibility of the
antecedent’s realization wrong24? It is finally time to face these questions that were
already implied in section 5.2.2.

22
Asking could there have been another person as Prime Minister is not the same thing asking: ”what if
there had been a another person such as Churchland as Prime Minister. What would have happened
then?” The second question concerns the causal role of Neville Chamberlain in the things that happened
after the Munich Agreement: was the person of Neville Chamberlain a causal difference-maker for those
things that followed or would the things have turned out the same way they did if some other person like
Churchland had been the Prime Minister?
23
I do not believe that the authors of the volume Counterfactual Thought Experiments in World Politics
ponder the conditions of realization of their counterfactual antecedents merely because the editors of the
volume Philip E. Tetlock and Aaron Belkin have asked them to do so (see Tetlock & Belkin 1996:18).
Instead, I suggest that there are reasons for presenting the questions on the plausibility or conditions of
realization of the antecedent, but these reasons follow from the rationale of causal reasoning.
24
It is not clear what historians take the relationship between causal explanation and the analysis on how
easily the antecedent could have realized to be. For instance, Tetlock & Belkin (1996: 7–9), Fearon
(1996: 41) and Lebow (2010: 45) make the remark that historians that work with single cases want to use
“plausible or conceivable causes” whereas more theoretically oriented scholars are more easily to accept
“miracle causes”, but they do not provide any detailed analysis what is the relation of plausible causes or
miracle causes to causal explanation. Julian Reiss is more explicit about this: the main claim of his article
is that historians’ way of thinking about causes differs from philosopher’s way of thinking about causes
as miracles. In Reiss’s view historians have their own method of causal inference that does not follow the
counterfactual theories of causal explanation that philosophers (Reiss mentions David Lewis) and some
others like the sociologist Max Weber have advanced.

80
 I argue that there are reasons to try to strive to fulfill the condition 4 in one’s
counterfactual inferences that aim at causal explanations. First, let us go back to the
basics: the aim of causal counterfactual inferences is to try to “explore” the causal
relation between A and B which in historians case means quite often making sense of
the vast network of causal links between A and B. What this means is that the center of
attention should be “the stuff” that is between A and B – the way to put the relationship
between A and B into focus is to isolate that relationship from other factors, including
the actual causal history of A. As already seen, the other conditions 1–3 of ideal
interventions also aim at isolating the relation between A and B. The conditions 1–3 are
means to pin point the disturbances that non-ideal interventions bring with them: the
ideal situation to assess the relation between A and B would be the situation where these
disturbances were not present or removed. The point of the condition 4 is similar: if we
want to see whether the values of B are dependent on the values of A, then the actual
causal history of A seems irrelevant and should, therefore, be “put aside” with an ideal
miracle intervention.
Secondly, the idea that interventions work like miracles saves us from the
debates on what point of history this intervention could have occurred, could it have
occurred at all and so on. For instance, in the case of counterfactual (5.1) where the
antecedent described Romania as s superpower, the non-interventionist methodology
was to try to rewrite history in a way that Romania would be superpower in the 1970’s.
Choosing the method of trying to rewrite history so that the antecedent is realized in a
plausible way leads to a series of difficult questions. At what time of history should we
start this rewrite and on what basis? At the turn of the 20th century? After the Second
World War? At the birth of Cleopatra and her nose?
Even if we reached an agreement on the exact point where the rewrite should
start, we would be faced with another problem: rewriting even small bits of history, not
to talk of decades or centuries, increases the complexity of our causal inferences
immensely, and easily we end up in a situation where we are unable to say what would
follow from the rewrites. That is why it is better to have one tactical “miracle” or
“ideal” intervention: ideal interventions aim at altering as little in the system as possible
so that the complexity of the system would not increase immensely. We try to keep the
system as simple as possible so that we do not end up at a loss with the complexity of
our counterfactual inferences. The Maxim machine gun intervention on the battle of
Waterloo in the middle of the battle alters the system so little that we can infer what

81
would have resulted: Napoleon would have won. Instead, changing the history so that
Maxim machine guns would have actually existed, increases the complexity of the
networks of causal connections to a point that we are no longer certain whether the
whole Battle of Waterloo could have occurred at all. Counterfactual inferences are hard
even with “miracle” interventions, and thus, there is no reason to make them even
harder with rewrites.
Thirdly, the idea that could-questions and could-counterfactuals should be used
as a criterion in selecting causes for events is misguided. Let us take an example of a
simple game with cards. In the game there are cards of four different colors: blue,
yellow, green and red. The rules are the following: a blue card can only be followed by
a yellow card, the yellow card with a green card and the green card by a red card and
red one by a blue one and so on. Let us presume that the game starts with a blue card
that is placed on the table. This means that necessarily the third card placed on the table
will be green and the fourth one red. Since we know the rules of the game we are
justified in saying that the green card on the table caused the placing of the red card on
the table and to make the following counterfactual inference: “if the card had not been
green, the next card placed on the table would not have been red”.
The fact we know that it was necessary that the third card was green since the
game started with a blue card (and that it was impossible for the third card to be of any
other color) does not prevent us from making the above-mentioned causal and
counterfactual inferences. Why not? That is because our causal judgments are informed
by knowing the rules of the game, not our knowledge on what was necessary, possible
or impossible given the causal history of the system. Knowing the rules allows to
imagine interventions that place a blue, red or yellow card in place of the green one and
to infer that in these cases the next card would not have been red. This counterfactual
inference helps us to conclude that it was the green card that was the cause of the
placing of the red card. The fact that it was not “objectively” or “actually” possible that
the green card could have been of different color does not prevent us from making the
causal claim that the green card was the cause of the red card.
In the same way, the knowledge that Neville Chamberlain’s hands were
politically tied and that it might have been impossible for him to act differently in
Munich and to abandon the appeasement policy, does not prevent us from making
causal and counterfactual inferences on the consequences of the appeasement policy. If
we have evidence that the Nazis would have changed their military strategy if

82
Chamberlain had been tougher in Munich, we have a reason to think that Chamberlain’s
behavior causally influenced the beginning of the Second World War. This evidence
gives as a reason to state the counterfactual antecedent “If Chamberlain had acted
differently”. The antecedent makes sense as a part of an explanatory causal claim
independently of the fact whether Chamberlain could actually have acted differently or
not. Knowledge on what was actually possible and could have happened does not affect
our causal inferences on what would have happened. The questions on what could have
happened and what would have needed to be different so that a certain antecedent could
have realized (could-counterfactuals) should be kept separate from the causal questions
on what would have happened if this had happened (would-counterfactuals).
Why are could-counterfactuals so popular among historians? Could-inferences
are used to answer two kinds of questions that seem to interest historians and general
public alike. First of all, there is the question of prevention: could a certain event or an
outcome that we wish not to have happened been prevented? Could the First World War
been easily prevented? Can we ensure that this thing does not happen again? What kind
of things we need to change in order to prevent these kinds of things from happening
again? To answer these questions we need to have an account of the causes of the event,
but in addition to the causal knowledge provided by would-inferences, could-inferences
are used to estimate how easily these causes could have been changed so that the
unwanted outcome would not occurred.
The questions of prevention lead quite often to questions of responsibility, guilt,
praise and blame25. The questions on whether certain actors could have done to
something to prevent the bad things from happening are often raised. Was Neville
Chamberlain in a position to break free from the appeasement policy and even to
prevent the Second World War from breaking out? If Neville Chamberlain was in a
position where he actually could not break away from the appeasement policy, he is not
to be blamed for following that policy as much as he would be if he had been totally
free to choose his strategy in Munich. Since the blaming and praising of historical actors
is a common phenomenon in the public discussion in many modern societies, historians
can contribute to this discussion by trying to deliver to the public the best
25
The reader might remember from the chapter 2 that also Daniel Nolan (2001: 15-17) took notice that
historians use counterfactuals to deal with the questions of moral responsibility. The theme of moral
responsibility was also discussed in the chapter four where it was argued that a historian should try to
keep the roles of researcher interested in causal explanations separate from role of the judge. This section
has showed that playing the role of the judge by using could-counterfactuals might be confusing from the
point of view of causal explanation.

83
historiographical knowledge that exists on the conditions that prevailed when an actor
did what he did. Knowledge on the conditions that prevailed helps to answer questions
whether the actor could have acted differently.
Like in the case of prevention, raising the finger of praise or blame often
requires that one has causal ideas that the actions of the person to be blamed or praised
actually contributed to the occurrence of a result that is considered either good or bad
(like the beginning of the Second World War). Therefore, blaming or praising Neville
Chamberlain for the appeasement policy would makes sense only when we are
convinced that the appeasement policy played a role in the beginning of the Second
World War26. So before historians get mixed with the issues of moral blame and praise
they should have at least some kind of knowledge of the causal relations that
prevailed27.
Claims that a certain person was the only person who was able to bring about a
certain result (that we attribute a moral value to) provide another interesting case of
attribution of moral praise. For instance, there are arguments that the actions of the
Athenian military leader Themistocles not only led to the winning of the Battle of
Salamis 480 BC, but that he was the only Greek military leader to be able to plan the
victorious tactics against the Persians (Hanson 2006). If Themistocles’ tactics lead to
the Greeks’ victory against the Persians, he is praised. If Themistocles was the only one
cunning enough to design the tactics that he did, he is praised even more, since this
means that no one else would have been able to save the Greek culture and the seeds of
Western culture – a result that the modern westerners often value.
The arguments that a certain result would have not happened without certain
acts or events are ubiquitous in historiography. What kind of counterfactual thinking are
these kinds of claims based on? On what basis can one make the argument that
Themistocles was the only one who could have won the Battle of Salamis? This is the
topic that I am going to return to in the next chapter where I discuss the various uses of
interventionist counterfactuals.

26
When people blame or praise historical actors, usually some kind of consequentialism prevails: actions
are deemed good or bad based on their consequences.
27
The could-question on possibilities and the would-question on causal dependence are usually presented
together. For instance, in economic history there has been a huge debate on the questions whether the
recent African history could have taken another course than the actual one, and what would the
consequences have been for the modernization and the economic development of the continent (Elster
1978: 192–196). The way these questions are often presented and investigated together probably makes it
difficult for practicing historians to keep the related counterfactuals separate: could-questions require
backward-looking counterfactual inferences, whereas would-questions do not.

84
6!!
!!!!!!!
Uses!of!interventionist!counterfactuals!
!!!!
Is it legitimate for the historian to consider alternative possibilities to events which have
happened?... To say that the thing happened the way it did, is not at all illuminating. We
can understand the significance of what did happen only if we contrast it with what
might have happened.
– Morris Raphael Cohen

In the previous chapter we saw that historians employ two kinds of counterfactual
inferences: firstly, they use backward-looking counterfactuals to infer things that could
have happened and secondly, they use (or as I claim they should use) interventionist
miracle counterfactuals to infer causal dependency relations. In this chapter, I will
discuss the most important ways in which the interventionist counterfactuals are put to
use in historians’ causal reasoning. I will start with separating four forms of of causal
weighting, and to conclude, I will briefly discuss how causal reasoning and
counterfactual inferences work in the construction of historiographical narratives.

6.1!Causal!weighting!
The metaphors of “causal weighting” and “weighting of causes” reveal one crucial role
that counterfactuals play in historian’s causal reasoning. E.H. Carr (1964: 90) once
claimed: ”Every historical argument revolves around the question of priority of causes”.
Historians are not merely interested in finding the causes of events, but they are also
keen on comparing the suggested causes in order to assign relative importance to them:
was this factor more important than the other causes in bringing about the effect or not?
This section discusses the various ways in which historians “weight causes” with the
help of counterfactual inferences in order to reveal their “significance”, “importance” or
“meaning” for the explanans1.

1
Historians talk very often of ”significance”, ”importance” and ”meaning” of causes, but as Martin
(1989: 53) notes philosophers of historiography have not paid too much attention to the question of causal
weighting, although the theme has been raised many times in the context of counterfactual thinking.
Martin , as well as de May & Weber (2003) and Robert Northcott (2008), all seek to solve the problem of
explanatory weighting with the help of contrastive explananda. Maw Weber (1949) discusses the
attributions of causal significance and “adequate causation” based on the considerations of necessity and
contingency that are reached with counterfactual inferences. None of the analyses that I am aware of
analyses the phenomenon of causal weighting as composing of several ways of inferring with the help of
interventionist counterfactuals, as it is presented in this section. It seems that the presupposition has been
that “causal weighting” is a monolithic activity composing of one type of questions and one type of
counterfactual inference.

85
6.1.1!Selecting!causes!and!fixing!the!causal!background!!!
The first way to employ interventionist counterfactuals to do causal weighting is to
separate causes from the background factors. This is the question that was discussed in
the two previous chapters. In chapter four it was pointed out that presenting well-
specified contrastive explanatory questions helps one to determine which factors to
include in one’s explanations as causes accounting for the contrast and what to leave out
as background conditions. This means that the factors that do not answer the specific
contrastive explanandum should be treated as background factors. The examples of why
Kennedy won instead of Nixon and why the U.S. election system is the way it is
showed clearly how the causes can be separated from the background factors.
Background factors explain different contrastive questions than causes. In other words,
background factors are the factors that do not make the difference presented in the
contrastive explanandum.
In chapter five we saw that among the possible causes that explain a certain
contrasts, some causes are believed to be more sensible than others. The considerations
on what to count as sensible causes and what not can also be said to be a way of
separating background factors from causes. The causes that we judge as insensible can
be treated as background factors: there is no reason to cite the lack of modern machine
guns or meteorite strikes as causes for the defeat at Waterloo, although these factors do,
in fact, account for the difference presented in the contrastive explanandum.
In addition, chapter five dealt with the difficulties that historians face in their
causal inferences when determining causes. The most important problems of causal and
counterfactual inferences in historiography stem from two sources. First of all,
historians often lack evidence that could support counterfactual and causal inferences.
Secondly, historians often deal with complex causal systems where there are so many
interacting variables and where also the “rules”, according to which the variables
interact, can change when the system is intervened on. This means that it is hard to do
“surgical manipulations” on the system and to change only the values of the variable of
interest, which would make counterfactual inferring neat and simple. Historians do not
often have enough evidence to infer what the non-ideal manipulations that they are able
to perform would do in the complex causal system of interest: would the intervention
that miraculously changes a to a* bring about the change in B, and if so, what kind of
change?

86
6.1.2!Using!various!contrastive!explananda((
Let us leave these issues of separating the background from the causes and problems of
inferring the right causes behind us, and assume a situation where a historian has plenty
of hypotheses on the causes of a certain outcome. This is, in fact, quite often the case.
Historians have plenty of causes that they believe to have contributed to bringing about
a certain outcome. The problem then is how to do causal weighting among these. In
other words, historians are not merely interested in knowing whether the facts a, c, d
and e were together sufficient to bring about the value b in B. They want to ask more
specific what if-questions on the causal roles that a, c, d and e played in bringing about
b: if we change c to c*, what kind of change will it bring about in B. Is it so that that no
event like b occurs, or does b occur altered into b*? This means presenting more
specific contrastive explananda than just asking why b occurred instead of not
occurring. Historians want to know whether all the factors a, c, d and e accounted for
the specific timing of or to a specific feature of b or was it so that only some of these
factors accounted for the various nuances in b.
Presenting various contrastive questions of a certain outcome is, therefore, one
way of doing weighing among causes. It is a way of finding out what specific aspects of
the outcome certain causes accounted for. The example of the Industrial revolution
presented in chapter four is an example of the use of this kind of methodology of
presenting various explananda of a certain outcome In addition to the method of
seeking causes to various aspects of the same outcome there are two even more specific
means of doing causal weighting: the analysis of contingency and the analysis of
sensitivity.

6.1.3!Analyses!on!contingency!and!overdetermination!
In the end of the fifth chapter, I introduced the following claim (6.1): Themistocles was
the only Greek leader that was able to beat the Persians in the Battle of Salamis. What
kind of counterfactual claim is this and how does it differ from the claim (6.2):
Themistocles caused the Greeks’ victory? Hanson (2006) who discusses the first claim
(6.1) supports it with the following evidence: there is evidence that all the other Greek
leaders of that time would not have made the similar kind of plan that Themistocles did.
The other leaders would have prepared tactics that would have likely resulted in a loss.
This evidence provides support for the counterfactual: “if Themistocles had been

87
replaced (by a miracle intervention) by some other Greek military leader of the time, the
Greeks would have lost the Battle of Salamis”.
The causal claim (6.2) that it was Themistocles (or to be more precise his
tactical plan) that caused the victory is, on the other hand, supported by the evidence on
the conditions that prevailed that day, detailed knowledge of the Persian and Greek
fleets, their sizes and normal tactics. This evidence helps to conclude that it was the way
that Themistocles decided to use the Greek fleet that caused the victory. For instance, if
the Battle had been fought at open sea instead of in the narrows of the channel of
Salamis, the Persians would have lost.
The claim (6.1) that Themistocles was the only Greek leader who was able to
win presupposes the causal claim (6.2) and the related counterfactual inferences. Before
one can say that Themistocles was the only man able to win, one needs to have evidence
that it was Themistocles in the first place who caused the victory. The counterfactuals
that support the claim (6.1) that it was Themistocles alone who was able to win the
Persians are therefore different from the counterfactuals that are used to define causes.
Causal inference makes use of interventionist counterfactuals that “test” on the basis of
evidence whether there is a right kind of dependency relation between a and b. Claims
like (6.1) are, on the other hand, used to raise the question whether there were other
causal routes that would have realized if the fact [a] had been removed altogether and
whether this would have led to an outcome that would have been somewhat similar to
[b]. Answering this kind of question requires evidence on two things. Firstly, one needs
to have evidence that there were other causal routes that would have realized if a was
removed. Secondly, one needs to know what would have resulted from the realization of
these other routes. In other words, one wants to know whether the actual route of where
a[a*] caused b[b*], was the only route that could have led to an outcome similar to b2.
When one removes the actual route of causes and analyzes the alternative routes
of causes that would have resulted from removing the actual route and tries to define
whether the outcomes of these routes would have been similar or not to b, one analyzes
the degree of contingency of the outcome b. If there is evidence suggesting that
removing the actual chain of causes was not likely to lead to developments that would
have resulted in an outcome similar to b, b is judged as contingent. If it seems on the

2
Philosophically, this kind of inference leads to difficult questions on the identity of facts or events: on
the basis of which criteria something is judged to be similar to b? But since historians seem to operate
with our unexplicated notions of similarity when answering these questions, I won’t get mixed with the
issue what is really means to say that x is similar to y.

88
basis of evidence that the removal of the actual chain of causes had led to developments
that would have resulted in an outcome similar to b, b is said to have been necessary or
overdetermined. It is important to notice that this kind of empirical analysis on the
degree of contingency of an outcome is far removed from the metaphysical questions of
necessity and contingency to which some antagonists of counterfactuals claim
counterfactual thinking to lead.
In order to illustrate this type of analysis a bit more, let’s discuss the breaking
out of the First World War – a theme that has attracted a lot of counterfactual “iffing”
that has aimed at revealing whether the war was necessary or not3. The discussion has
revolved around the question: if there had been no Sarajevo shots, would a war similar
to the First World War have broken out sometime after July 1914 or in the years after
1914? The idea of this question is to weigh the importance of two factors that we
believe to have contributed to the breaking out of the First World War. We believe that
the alliances that the European superpowers formed in the late 19th and early 20th
centuries played an important role in the breaking out of the war: without this kind of
alliance structure the Sarajevo shots, which we regard as the other important cause of
the war, would not caused a turmoil that dragged so many European countries to war.
So, no war due to the Sarajevo shots without the alliances. This is the inference that
most historians can agree on. What is the source of much more debate is the question
whether there would have been a war without the Sarajevo shots. In other words, among
the actual causes of the war, were the alliances more important than the Sarajevo shots
in bringing about the war? To answer this question we have to find evidence that could
be used to conclude what would have happened if the Sarajevo shots had not been shot
or had been removed by a miracle intervention: are there reasons to believe that
removing the shots had resulted in alternative causal routes that would have lead to a
similar kind of war?
How does this kind of analysis of contingency and overdetermination relate to
the classic question of overdetermination in philosophy? Overdetermined causal
systems have posed a vast challenge for philosophers defending counterfactual theories
of causation, since simple counterfactuals of the form “if a had not occurred, b would
not have occurred” fail to capture the causal relations in overdetermined systems. In
cases of overdetermination there are two (or more) causes, a and d, present that both

3
For a nice summary of the discussion, see Lebow 2010: ch. 3.

89
will lead to the occurrence of b[b*] independently of each other. The counterfactual “if
a had not occurred b would not have occurred” is false since if a had not occurred, d
would have still caused b[b*]. What is a troubling technical worry for a philosopher4, is
a remarkable empirical question for a historian, however. If a historian manages to find
evidence for the claim that the beginning of the First World War was overdetermined,
she is entitled to make various inferences on the beginning of the war. She can, for
instance, claim that the alliances were a more important causal factor for the beginning
of the war than the Sarajevo shots, since even without the shots the alliances formed an
explosive cocktail that was ready to explode if only it was given a spark like the
Sarajevo shots. She can also conclude that the shots can be used to answer the question
why the war broke out on July 1914 and not some other time, but that the shots do not
provide a sufficient answer to the question: why did the war break out instead of not
breaking out at all? In order to account for the latter contrastive questions she needs to
cite other causes5.
The analyses on contingency are a form of causal weighting where a historian
tries to determine with the help of interventionist counterfactuals what factors in the
causal chain of events leading to an outcome were more important than others. Was the
context or certain factors of the chain of events causally significant for a certain result
so that removing a part of the causal chain or the whole chain6 would not have mattered
for the outcome?7 The analysis of contingency is about removing the actual chain of
causes and seeing where the interactions between other factors (the background factors)

4
Woodward (2003: 77–79) claims to have solved the problem of overdetermination with his
interventionist account. Interventionist counterfactuals manage to capture the right causal structure in
cases of overdetermination: if b is overdetermined by a and d and if a is the actual cause of b,
interventionist counterfactuals lead to the result that it is a that causes b, not d. Simple non-interventionist
counterfactuals, on the other hand, would lead to the wrong conclusion: a is not the cause of b.
5
If on the other hand, the historian would find evidence to conclude that the Sarajevo shots were the only
causal route that leading to war, she is entitled to attach much more causal importance to the shots: the
shots not only explain the timing of the war, but also are an important factor in explaining why the war
broke out in the first place. In this case the beginning of the war would seem highly contingent.
6
It depends on the case under study whether is it more feasible to talk of removing a part of the chain
rather than removing the whole chain. In the case of Themistocles and the Battle of Salamis it is more
natural to talk of removing the whole causal chain, since there the intervention on the leader of the battle
means changing the whole course of the battle. In the case of the First World War, removing the Sarajevo
shots and the things that followed the shots is more naturally described as removing one part of the causal
chain that lead to the First World War.
7
Analyses on contingency and overdetermination are also ways to determine causal turning points in
history. Turning points are factors that are difference-makers to various events that succeeded them: if we
remove or change the turning point in some way, it would result into massive changes. For instance, it has
been clamed that if the Greeks had lost the battle of Salamis, the Western culture had not developed at all
(Hanson 2006; Strauss 2006). The birth and rise of the Western culture are thus seen as highly contingent:
there is one single event, the victory of the Battle of Salamis that the Western culture was depended on. If
this event had been removed, the Western culture would not have developed.

90
would have led after the removal. The fourth form of causal weighting – sensitivity
analysis – that I will discuss next is the mirror image of this kind of inference: the point
of sensitivity analysis is to change (or remove) some of the other factors in order to see
whether the actual chain of causation would have stayed the same.

6.1.4!Analyses!on!sensitivity!
Counterfactual dependence and invariance are not the only significant properties that
causal relations have. Causal relations and the related counterfactuals also have the
feature that can be called ‘sensitivity’ (Woodward 2006: 2; see also: Ylikoski &
Kuorikoski 2010). A causal claim and the related counterfactuals are sensitive if they
hold in the actual circumstances, but fail to hold in circumstances that depart in various
ways from the actual circumstances (Woodward 2006: 2). To the extent to which a
causal claim and the related counterfactual would fail to hold under various changes in
the actual circumstances, defines how sensitive or insensitive the causal claim and the
related counterfactuals are.
For instance, pushing the break when driving a car makes it stop. This causal
claim is insensitive to various changes in the background conditions. For instance, the
speed of driving does not affect the whether the car eventually stops or not (it, of
course, affects the time it takes for the car to stop). It does not matter whether I am
driving in rain, on a German highway or with my high heel shoes on for the car to stop
or not. The causal relation is, on the other hand, sensitive to changes in the break system
of the car: if oil gets into the break system, the car might fail to stop.
Sensitivity analysis is a form of causal reasoning that employs interventions and
counterfactual inferences: one tries to find evidence to support one’s inferences on what
would happen to the causal relation of interest under various interventionist changes in
the circumstances that actually occurred. Successful sensitivity analysis contributes to
our causal knowledge by further expanding the area of what if-inferences we can make:
it gives us the ability to answer correctly not only what if-questions on the invariance of
the causal relation but to what if-questions on the holding of this invariant relation in
changing circumstances. Sensitivity analysis can, therefore, be seen as a form of causal
weighting: how important was the fact that the other factors took the values they
actually took for the causal relation to hold and to produce the outcome it did? In other

91
words, how did the actual circumstances causally contribute to the fact that the causes
actually were able to produce the effect?8
In historiography, there are several examples that can be interpreted as
sensitivity analysis. To further expand Hanson’s (2006) case study on the Battle of
Salamis let us bring up the following causal questions: What was the importance of the
weather conditions being the way they were or the Persians ships being the way the
were or the color of Themistocles shoes for the holding of the causal relationship that
brought about the victory of the Greeks? Would Themistocles’ plan have worked if the
color of his sandals had been different? Would it have worked if a storm had struck in
the middle of the battle or if the Persian boats had been of similar size to Greek boats?
What if the “war hawk” governor-general of Bosnia-Herzegovina, Oskar Potiorek, had
been killed in Sarajevo in 1914 instead of the more peace-loving idealist Franz
Ferdinand? Would this have made a difference in the causal process that lead to the
First World War (see Lebow 2010: 83–84)?
What can historians achieve with these kinds of questions and sensitivity
analyses? First of all, knowing what kind of factors the holding of the actual causal
relation was dependent on helps us to make informed decisions on what kind of factors
to ignore as irrelevant background factors to be excluded from the historians’
descriptions of causal processes. If we are entitled to infer that an intervention on the
color of Themistocles’ shoes right before or in the middle of the Battle of Salamis, does
not make a difference to the carrying out or succeeding of Themistocles’ plan, we have
a good reason not to include Themistocles’ shoes into our descriptions on the causal
processes on the victory. The weather conditions and the size of the Persian boats, on
the other hand, seem to be background factors that needed to be quite close to their
actual values in order for Themistocles’ plan to work. Therefore, sensitivity analysis
helps to determine the factors that were irrelevant to the success of Themistocles’ plan
and the factors contributed to the success of Themistocles’ plan. This way our picture of
the causal processes governing the Greek victory becomes “thicker” and fuller: we do

8
Sensitivity analysis can also be regarded as a way to make inferences on the contingency or necessity of
a result (see Ben Menahem 2011 for further discussion on this). If the causal relation between a and b is
extremely sensitive, we usually consider b to be highly contingent. If, on the other hand, the causal
relation is very insensitive to changes in the background conditions, b is regarded as a necessary result:
we would have needed a lot more interventions (or one “massive” intervention) on the background factors
in order to stop b from occurring. The analysis on contingency and necessity of events that makes use of
sensitivity analysis is not the same thing as the analysis of contingency and overdetermination presented
in the previous section. Sensitivity analysis concerns changing the background factors and analysis of
overdetermination works by changing or removing part of the causal chain of interest.

92
not merely causally understand the causal relation, but also the relevant background
conditions that explain why the causal relation held. This way we also get information
on what factors one should have intervened on in order to prevent the causal process
and the outcome of the process from occurring – interesting information for historians
who are also concerned about questions on how a certain unpleasant result could have
been avoided9?
It seems that many causal relations that historians deal with are relatively
sensitive: there are usually quite many interventions on the actual values of several
variables that could have affected the causal relation. This kind of sensitivity of the
causal relation also blurs the identity of the causal relation itself: should we count the
dozens or hundreds of “background conditions” that do make a difference as
endogenous parts of the causal relation or not? I do not have any general recipe to solve
this question: the issue needs to be settled case-by-case and sometimes more detailed
contrastive questions might help in cutting down the abundance of causally relevant
factors.
As counterfactual inferences on invariance, the analyses of sensitivity
sometimes turn out to be extremely challenging. What makes counterfactual inferences
on sensitivity sometimes difficult are, however, the same problems of complexity and
modularity that make counterfactuals on invariance difficult. If the network of causal
relations that is of interest to a historian is complex in the way that it involves dozens of
variables most of which interact with each other, counterfactual inference becomes
difficult. For instance, should we interpret the intervention on the weather during the
Battle of the Salamis in the way that this intervention would not have affected the
carrying out of Themistocles’ plan? Are there reasons to assume that Themistocle’s
would have reacted to a sudden change in weather by revising his plan altogether? If
Themistocles had revised his plan, what would have resulted? Do we have enough
evidence to know that? The biggest problem for historian’s causal inference is not
whether a certain intervention could have realistically occurred or not. The biggest

9
For instance in his discussion of the First World War, Richard Ned Lebow asks whether the First World
War could have been avoided or not? This question is intriguing for a historian as an intellectual
challenge, but also important for the general audience and politicians concerned with the issues of
responsibility. In order to answer this question one needs, first of all, to determine with the help of
interventionist counterfactuals the factors that the holding causal relation was depended on. After that one
employs could-counterfactuals to determine how easily the factors on which the holding of the causal
relation was dependent could have been changed.

93
problem is that historians sometimes do not have enough evidence to infer what would
have happened in the network of causal relations after an intervention.
The problem of defining what would actually happen after an intervention is not a
problem that only historians face. James Bogen (2004: 5–6; 2004: 5–6; see also
Hausman and Woodward 1999: 537, n.12) introduces an example from astronomy:

Suppose the mutual gravitational attraction of the heavenly bodies makes it

impossible for anything to change the moon’s distance from the earth enough to
increase or to diminish its influence on the tides without changing the positions of
other heavenly bodies so that their influence on the tides changes significantly.
Then if something moved the moon closer or farther from the earth, the ensuing
value of the tides would reflect the causal influence the other heavenly bodies
exert independently of the moon’s influence on the tides and moving the moon
would not qualify as an intervention on the moon’s distance from the earth with
respect to its tides.
(Bogen 2004: 5–6.)

Ideal interventions are hard to make in systems that have multiple interacting parts,
since changing one part would affect other parts in a way that it becomes hard to
estimate the end result. The contrastive counterfactual theory does not help to solve the
problem of how to deal with complex systems where an intervention affects not just the
variable, but several variables through the interactions between these variables. It seems
to be a task that is left to the practicing scientists and historians to figure out what kind
of and how much of causal knowledge can be gained from the actual causal systems that
do allow clean-cut surgical interventions. However, this does not render the contrastive
counterfactual theory useless. The theory still provides guidelines where to move in our
causal inferences: what kind of questions to ask and what kind of experiments to design
and what kind of evidence to look for and use to support causal claims. As seen in the
chapters in this thesis, this is quite a contribution. Before closing this chapter let us take
a look at one more implication of this theory and explore the relationship between
counterfactual inferences and historians’ main representative means – the narrative.

6.2!Explaining!with!narratives!
As the previous sections have shown historians usually need to deal with a variety of
causes when constructing causal explanations. E.H Carr makes a point by claiming
(1964: 89): “[t]he first characteristic of the historian’s approach to the problem of cause
is that he will commonly assign several causes to the same event.” Dealing with several

94
causes leads to the problems of causal weighting described in the previous sections, but
it also creates another problem: how to present an abundance of causes in a cognitively
salient way? Historians have solved this problem by making use of the distinct human
capacity to tell stories, and thus narratives have become the main tool of representation
in historiography.
Originally historiographical narratives resembled stories in many respects: they
presented in chronological order how certain events interacted together to produce a
certain outcome. In current historiography, narratives have become more versatile: the
causes of events and consequences are not necessarily presented in chronological order
(Fellman & Rahikainen 2012: 9–11). Instead, more thematical approaches are preferred.
The causes might be presented grouped, for example, in structural and individual level
causes, and what is more important, chronological order is often broken with
considerations of the significance or importance of causes.10 Despite these changes, the
central aim of the narrative style of representation has stayed the same: the aim is to
causally explain how the event of interest came about11. The last section of the thesis
deals with the question how counterfactual inference contributes to narrativist causal
explanation.

6.2.1!Counterfactual!inference!and!the!construction!of!narratives!
There has been a long-standing debate in philosophy of history from the 1960’s
onwards on the question whether ”narrative explanation” is an independent form of
explanation and whether narratives explain in virtue of being ”narratives” (see e.g. Dray
1971; Carr 2008; Froeyman 2009). The claim of this section is that narrative
explanation is not an independent form of explanation and that narratives are not self-
explanatory. The question is therefore the following: when a historian works with an
abundance of suggested causes in order to construct a narrative, how does
counterfactual reasoning work and how the reasoning contribute to the fact that
narratives explain?
Let us start from the problem of explanatory selection that was already touched
upon in the chapter four in section 4.2 that discussed the role of contrastive explananda
in selecting explanatory causes. The issue can be further illustrated with another

10
Sometimes the considerations of significance are presented by means of explicitly presented
counterfactuals.
11
Causal explanatory questions can also have the form of “how”-questions: how did b come about instead
of b*?

95
factors came together in the process that led to war. In the end, however, counterfactual
inferences play a part in the selection of causes and explanatory factors. The historian
weighs causes in order to decide what factors are more important than others and these
considerations play a role in what the historian leaves out and how she presents the
factors that she includes in the narrative. The central claim of this thesis has been that
the various forms of causal thinking that historians engage in are also forms of
counterfactual thinking. The selection of causes, analyses of contingency and sensitivity
all make use of interventionist counterfactuals and all these can play a role in the
construction of narratives.
In the light of the contrastive counterfactual theory, narratives depict chains of
causation that are believed to have resulted in the outcome(s) of interest. In other words,
narratives aim at depicting the factors that are considered on the basis of counterfactual
inferences to have been the difference makers to a certain outcome. This is the feature
that separates narratives from mere chronicles that present events followed by other
events. Chronicles are descriptions of the form “this event was followed by another and
after that something else happened”. Narratives, instead, are causal accounts – usually
full of causal verbs like led to or fostered – that aim at depicting chains and bundles of
causal relations that answer the contrastive question: Why this outcome instead of some
other kind of outcome? Among the causes that answer this contrastive question one can
also do further causal weighing. For instance, if one concludes that the conservative
desires to safeguard status quo in Austria, the belief that the war is inevitable among the
German military officers, as well as poor Russian crisis management (see table 6.1)
were all causes that played a role in the beginning of the First World War, one can ask
further: What kind of role? What kind of more specific contrastive questions can these
factors answer? Are some of these factors more important than others in explaining a
certain contrasts? These are all forms of causal reasoning that affect the construction of
narratives. The conclusion of these considerations is that narratives do not explain in
virtue of being narratives, they explain in virtue of representing complex chains and
bundles of causal relations that are believed to support various counterfactual what if-
inferences.
However, the narrative form usually hides the what if-inferences that affect the
construction of narratives. The narratives themselves contain explicit counterfactual
claims only occasionally: their explicit use is usually confined to situations where a
historian discusses the causes she considers to be most significant ones. Hence,

97
counterfactual reasoning remains in many cases implicit, and it might be that the
historian does not pay too much attention to the inferences that guide her causal
thinking: causal and counterfactual thinking stays hidden behind the phrase “historian’s
imagination”.
There are dangers in this implicitness, however. For instance, implicit
counterfactual thinking might lead to situations where historians think that they have
more causal knowledge than they actually have: a coherent story of the way things
evolved might give us the impression that we know the causes of a certain event.
Asking counterfactual “what if”-questions might reveal this is not the case. When one
approaches evidence with the help of counterfactual questions, it might reveal that the
evidence is all but conclusive in determining what to count as causes and what not. The
explicit use of counterfactuals might prevent one from doing hasty causal conclusions,
and also give one strong reasons to believe in one’s causal conclusion: if one can
present the evidence that supports one’s counterfactual claims, the evidence is evidence
also for the related causal claim.
Therefore, there are better guides than the mere unarticulated power of
imagination to direct the construction of narratives. The explicit use of counterfactuals
and contrastive explanatory questions and various forms of counterfactual thinking
contribute to the quality of historiographical research – they force one to think more
explicitly the content of one’s explanatory causal claims and the evidence that one has
to support these claims. Therefore, I hope that, in the future, the narrative style will also
witness an increase in the explicit use of counterfactual inferences as well as an increase
in the evaluation of evidence in the light of these inferences.

98
7!
!
!
!!!!!!
Conclusions!

There is no a priori guarantee that every subject matter must be a suitable domain for
causal explanation.
– James Woodward (2003: 314)

The task set for this thesis in the end of the second chapter was to introduce a theory of
causal explanation that would fulfill two desiderata: firstly, the theory should explicate
the link between causal explanation and counterfactuals in order to assess the claims of
both the antagonists and the protagonists of counterfactuals. In other words the aim is to
provide a balanced solution to the debate between the two sides. Secondly, the theory
should not only explicate the relation between causal explanation and counterfactuals,
but also do it in a way that can be shown to be methodologically fruitful. These are the
two desiderata set from the side of philosophy of historiography and practicing
historians. From the point of view of philosophy of science the most important task is to
investigate whether the most distinguished theory of causal explanation of today applies
to historiography, and if it does how well? Now its time to summarize to what extent the
theory of causal explanation presented in this thesis – the contrastive counterfactual
theory – copes in fulfilling all these three tasks.
For the protagonists of counterfactuals the conclusions of the thesis are
soothing: there is a conceptual link between causal and counterfactual claims and the
contrastive counterfactual theory manages to explicate the link in a way that avoids the
most dangerous pitfalls of counterfactual thinking: the theory secures the asymmetry of
causal relations, does not let mere correlations to explain and does not allow
backtracking counterfactual inferences to count as causal inferences. The doubts that the
antagonists have presented can also be alleviated. Counterfactuals do not necessarily
lead into hard-core metaphysics. The notion of “possible worlds” is not needed at all in
the contrastive counterfactual theory, and even if one wanted to use the semantics of
possible worlds, the semantics could be treated as an inferential tool in a way that does
not involve any kind of ontological commitment. Neither does the use of
counterfactuals lure one into the metaphysics of modal notions such as necessity and
contingency: the questions of necessity and contingency can also be empirical questions
that require counterfactual inferences and empirical evidence to be answered.

99
 The accusations of loose semantics can also be dealt with. It is true that the
sentence “If Julius Caesar had been in charge of U.N. Forces during the Korean War, he
would have used (a) nuclear weapons or (b) catapults“ is murky. But as already stated in
chapter two the murkiness stems, in part, from the fact that based on the empirical
evidence we have no clear criteria to choose between the alternatives. Secondly, the
sentence is not a causal claim. When it comes to causal claims the truth makers of the
claims are objective invariant dependency relations. Singular causal claims that capture
objective invariant relations can be specified with the help of double contrasts ‘a [a*]
caused b [b*]’. The double contrast spells out the content of the causal claims in the
clearest way possible. The truth makers of causal claims are given with the help of
interventionist counterfactuals:

a[a*] causes b[b*] if and only if there is an intervention on the variable A such that:
if the intervention changed a to a*, this would bring about a change from b to b*.

As it was pointed out in the fifth chapter, not all interventionist counterfactuals that
fulfill the above condition are of interest to historians as causal explanations, however.
On the contrary, some claims citing causes seem to be quite implausible and insensible.
The obscurity of implausible counterfactuals can be alleviated with the help of the
existence and mechanistic criteria that explain why some causes are considered
implausible and some not. In the end, our considerations of implausible causes and
counterfactuals depend on our evidence and theoretical conceptions how the world
works, although the truth of interventionist counterfactuals depends entirely on whether
they track down objective invariant dependency relations.
How can we then know whether our counterfactual inferences are true or not, in
other words, whether they track down the objective invariant dependency relations?
This question conveys the epistemic concern of the antagonists of counterfactuals. It is a
good concern indeed. Although the principled claim that we cannot know the truth of
any counterfactual claim is too drastic, there are cases where determining the truth of
counterfactuals turns out to be guesswork. This can happen for two, often interrelated,
reasons: we might not have enough evidence to decide the truth of the counterfactual or
then our counterfactual inferences might become too complex when dealing with
complex causal systems that we usually know only in part. These are problems that
cannot be solved on the general level of philosophical theory; instead, these problems

100
are best dealt with case-by-case basis by trying to find new ways to handle existing
evidence, to generate evidence and to make inferences. As pointed out in the fifth
chapter, counterfactual inferences or “thought experiments” are not a way to generate
new evidence – their quality depends entirely on the quality of evidence and the way the
evidence is used.
The antagonists’ attitude towards counterfactuals seems to have been partly
motivated by disciplinary concerns and the desire to protect historiography’s identity
when faced with incursions from other disciplines. Like every discipline, historiography
has its unique methods of research, but from the point of view of causal explanation
historians’ explanatory reasoning can be said to resemble the causal reasoning that is
also employed in other fields of study. Historiography does not have its own ways of
causal reasoning and explanation – also narratives are constructed and assessed with the
help of counterfactual reasoning. In addition, opening up the disciplinary boundaries
and borrowing theories from other disciplines can in some cases be an apt way to get
new empirical evidence to support one’s causal inferences.
The doubts of the antagonists are therefore answered and the argumentative gaps
of the protagonists filled by the contrastive counterfactual theory. How about the
methodological insights that the theory was supposed to provide? Based on the
discussions of the previous chapters I have summarized the methodological guidelines
provided by the contrastive counterfactual theory into a toolkit of causal reasoning in
historiography:

1) When engaging in causal reasoning, specify the explanandum in contrastive

terms: why b and not b*? This clarifies your explanation-seeking question,
which often helps in i) picking the right kind of causes and separating those
causes from the causal background ii) turning big questions into smaller ones
that are easier to answer iii) weighting causes.

2) The reliability of causal claims and the related interventionist counterfactuals

depends on the strength of the evidence. Asking explicit what if-questions is an
apt way to assess your evidence: what causal claims does the evidence support
and what not?

101
 3) If you are engaged in causal weighting specify, whether you are a) separating
causes from the background factors b) doing causal weighting among causes c)
analyzing the contingency of an outcome d) analyzing the sensitivity of a causal
relation.

4) When engaging in counterfactual reasoning that aims to define causes of events,

counterfactual inferences should follow the guidelines characterized in the fifth
chapter (STEP1* – STEP4*). Be careful in trying to fulfill all the conditions of
ideal interventions. The conditions set the regulative ideal for our causal
inferences, but it might be that the requirements on modularity (e.g. condition 2)
or the requirement on the absence of correlations (condition 3) cannot easily be
fulfilled.

5) Could-counterfactuals that deal with possibilities should not be mixed with

would-counterfactuals that deal with causes. Could-inferences are useful in the
context of attributing moral responsibility and preventing unwanted outcomes,
but these questions should not be used to limit the area of sensible causal claims.
Could-counterfactuals are backward looking, whereas causal would-inferences
are not. Causal inferences concentrate on what happens in the causal system
after the “miracle” intervention; one does not need to worry about whether the
changed caused by the intervention was actually possible or not.

It seems that applying the contrastive counterfactual theory to historiography is a

methodologically fruitful endeavor that can give some concrete guidelines for causal
reasoning in historiography. How does the situation look like seen from the side of the
contrastive counterfactual theory and philosophy of science? Has the application of the
theory into historiography been a success? Is there still room for James Woodward’s
doubts? To begin with the contrastive counterfactual theory easily outflanks the most
discussed theory of explanation in the philosophy of historiography, namely the
covering-law model. Causal explanation of human action is not a problem for this
theory, either. In addition, the idea of contrastive explanandum seems to work in the
context of historiography in a nice way and to solve some big issues like the debate of
subjectivity of causal explanations. The idea that interventions work like miracles helps
to distinguish could-counterfactuals from causal would-counterfactual and to see that

102
causal explanation is not the only context where historians use counterfactuals. In
addition the theory helps to distinguish the different forms of causal weighing from each
other.
Historiography also seems to point out some of the biggest weaknesses of the
contrastive counterfactual theory. The implausible counterfactuals discussed by several
protagonists of counterfactuals such as the counterfactual on Napoleon and Maxim
machine guns reveals that not all causes are considered interesting or sensible. The
theory needs to be supplemented with some extra considerations to account for this.
Secondly, the idea that an intervention should directly affect only the purported
explanans variable in a causal system of interest runs into difficulties in the context of
historiography: if an ideal intervention is not possible on a system, the non-ideal
intervention seems at best to provide non-ideal causal knowledge. How to characterize
this non-ideal causal knowledge then? This is a question that remains to be answered. *

*
I would like to thank Petri Ylikoski for introducing me to the topic of the thesis and the relevant
literature. Jaakko Kuorikoski is praised for supervising my work and for his insightful and detailed
comments on the script along the way. The comments I received in the seminars of professors Gabriel
Sandu and Matti Sintonen helped me decide which direction to take in the early phases of the project. I
am also grateful to Piiku Jakola for the last-minute language check and to Ilmari Hirvonen for his support.
In addition, I will express my gratitude to the philosophy of science of community at University of
Helsinki for providing me with an inspiring place to work.

103
 References!
Achinstein,-Peter-(1983):-The$Nature$of$Explanation.$Oxford:-Oxford-University-Press.-
Aron,-Raymond-(1961):-Introduction$to$the$Philosophy$of$History,-translated-by-George-J.-Irwin.-
London:- Weidenfeld- and- Nicholson.- The- original- title- Introduction$ à$ la$ philosophie$ de$
l’histoire-published-1938-by-Librarie-Galimard.--
Aunesluoma,-Juhana-(2010):-”Teoria-ja-historioitsijan-valinnat”,-Kosmopolis$40,-2:-61–71.-
Bates,- Robert- H.;- Greif,- Avner;- Levi,- Margaret;- Rosethal,- Jean`Laurent- and- Weingast,- Barry--
(1998):-Analytic$Narratives.$New-Jersey:-Princeton-University-Press.--
Ben`Menahem,-Yemima-(2011):-”Historical-Necessity-and-Contingency”,-in-in-A$Companion$to$
the$ Philosophy$ of$ History$ and$ Historiography,- ed.- Tucker.- Paperback- edition- (hardback-
edition-published-2009).-West-Sussex:-Wiley`Blackwell.-120–130.-$----
Bogen,- Jim- (2004):- ”Analysing- Causality:- the- opposite- of- counterfactual- is- factual”,-
International$Studies$in$the$Philosophy$of$Science-18,-1:-3–26.-$-
Bulhof,-Johannes-(1999):-“What-if?-Modality-and-History”,-History$and$Theory-38:-145–168.-
Bunzl,- Martin- (2004):- ”Counterfactual- History:- A- User’s- Guide”,- American$ Historical$ Review$
109,-3:-845–858.-
Burton-Adams,-George-(1909):-“History-and-Philosophy-of-History”,-American$Historical$Review$
14:-221–236.-
Carr,-David-(2008):-“Narrative-Explanation-and-its-Malcontents”,-History$and$Theory$47:-19–30.--
Carr,-E.H.-(1964):-What$is$history?-Middlesex:-Penguin-Books.-
Cederman,-Lars`Erik-(1996):-“Rerunning-History-–-counterfactual-simulation-in-world-politics”,-
in-Counterfactual$Thought$Experiments$in$World$Politics,$ed.-Tetlock-&-Belkin.-Princeton:-
Princeton-University-Press.-248–267.-
Chisholm,-Roderick-M.-(1946):-“Contrary`to`fact`conditional”,-Mind$55,-220:-289`307.-
Climo,- T.A.- &- Howells- P.G.A.- (1976):- “Possible- Worlds- in- Historical- Explanation”,- History$ and$
Theory$-15,-1:-1–20.--
Cohen,- Raphael- Morris- (1964[1931]):- Reason$ and$ Nature:$ the$ Meaning$ of$ Scientific$ Method.$
Glencoe:-The-Free-Press.-
Collingwood,- R.G.- (1994[1946]):- The$ idea$ of$ history.- Revised- paperback- edition.- Oxford:-
Clarendon-Press.--
Craver,-Carl-(2007):-Explaining$the$Brain:$Mechanisms$and$the$Mosaic$Unity$of$Neuroscience.-
Oxford:-Clarendon-Press.-
Croce,-Benedetto-(1966):-“-‘Necessity’-in-history”,-in-Philosophy,$Poetry,$History:$An$Anthology$
of$Essays,$translated-by-Cecil-Sprigge.-London:-Oxford-University-Press.-
Crouzet,- Francois- (2001):- A$ History$ of$ the$ European$ Economy,$ 1000N2000.$ Virginia:- Virginia-
University-Press.-
Danto,-Arthur-(1965):-Analytical$Philosophy$of$History.$Cambridge:-Cambridge-University-Press.-
Davidson,- Donald- (1963):- ”Actions,- Reasons- and- Causes”,- The$ Journal$ of$ Philosophy- 60,- 23:-
685–700.--
De-May,-Tim-&-Weber,-Erik-(2003):-”Explanation-and-Thought-Experiments-in-History”,-History$
and$Theory$$42:-28–38.-
Donagan,- Alan- (1964):- “Historical- Explanation:- the- Popper`Hempel- Theory- Reconsidered”,-
History$and$Theory$4,-1:-3–26.-

104
D’Oro-Giuseppina-(2011):-“Historiographic-Understanding”,-in-A$Companion$to$the$Philosophy$
of$ History$ and$ Historiography,- ed.- Tucker.- Paperback- edition- (hardback- edition-
published-2009).-West-Sussex:-Wiley`Blackwell.-142–151.-$--
Dray,-William-(1962):-”Some-Causal-Accounts-of-the-American-Civil-War”,-Daedalus$91:-578-–-
592.--
---—-(1963):-“The-Historical-Explanation-of-Action-reconsidered”,-in-Philosophy$and$History,$ed.-
Hook.-New-York:-New-York-University-Press.-105–135.-
---—-(1971):-“On-the-Nature-and-Role-of-Narrative-in-Historiography”,-History$and$Theory$10,-2:-
153–171.----
---—- (1978):- “Concepts- of- Causation- in- A.J.P- Taylor’s- Account- of- the- Second- World- War”,-
History$and$Theory-17:-149–174.--
Elster,-Jon-(1978):-Logic$and$Society:$contradictions$and$possible$worlds.$New-York:-John-Wiley-
&-Sons.-
Fearon,-James-D.-(1991):-“Counterfactuals-and-Hypothesis-Testing-in-Political-Science”,-World$
Politics$43,-2:-169–195.-
---—- (1996):- “Causes- and- Counterfactuals- in- Social- Science:- exploring- an- analogy- between-
cellular- automata- and- historical- processes”,- in- Counterfactual$ Thought$ Experiments$ in$
World$Politics,$ed.-Tetlock-&-Belkin.-Princeton:-Princeton-University-Press.-39–67.-
Feinstein,- Charles- H.- &- Thomas,- Mark- (2002):- Making$ History$ Count:$ a$ primer$ in$ quantitive$
methods$for$historians.-Cambridge:-Cambridge-University-Press.-
Fellman,- Susanna- &- Rahikainen,- Marjatta- (2012):- “On- Historical- Writing- and- Evidence”,- in-
Historical$ Knowledge$ –$ in$ quest$ of$ theory,$ method$ and$ evidence,- ed.- Fellman- &-
Rahikainen.-Newcastle:-Cambridge-Scholars-Publishing.--5–44.--
Ferguson,- Niall- (2011[1997]):- “Introduction”,- in- Virtual$ History:$ alternatives$ and$
counterfactuals,-ed.-Niall-Ferguson.$London:-Penguin-Books.-1–90.--
Fischer,- David- (1971):- Historians$ Fallacies:$ toward$ a$ logic$ of$ historical$ thought.- London:-
Routledge-and-Kegan-Paul.-
Fogel,-Robert-(1964):-Railroads$and$American$economic$growth:-essays$in$econometric$history.$
Baltimore-&-London:-The-Johns-Hopkins-Press.-
Fogel,-Robert-&-Elton,-G.R.-(1983):-Which$road$to$the$past?$Two$views$of$History.$New-Haven-&-
London:-Yale-University-Press.-
Førland,-Tor-Egil-(2004):-”The-ideal-Explanatory-Text-in-History:-a-plea-for-ecumenism”,-History$
and$Theory-43,-3:-321–340.--
Froeyman,-Anton-(2009):-”Concepts-of-Causation-in-Historiography”,-Historical$Methods$42,-3:-
116–128.--
Garfinkel,- Alan- (1981):- Forms$ of$ Explanation$ –$ rethinking$ the$ questions$ in$ social$ theory.- New-
Haven-&-London:-Yale-University-Press.-
Glennan,- Stuart- (2010):- “Ephemeral- Mechanisms- and- Historical- Explanation”,- Erkenntnis$ 72:-
251–266.-
Gould,-J.-D.-(1969):-“Hypothetical-History”,-The$Economic$History$Review$22,-2:-195–207.--
Hanson,-Victor-Davis-(2006):-“A-Stillborn-West:-Themistocles-at-Salamis-480-BC”,-in-Unmaking$
the$West:$“WhatNif?”$Scenarios$that$rewrite$world$history,$ed.-Tetlock,-Lebow-&-Parker.-
Ann-Arbor:-The-University-of-Michigan-Press.-47–89.-
Hart,-H.L.A-and-Honoré,-Tony-(1985):--Causation$in$the$Law.-2nd-ed.-Oxford:-Clarendon-Press.-

105
Hausman,- Daniel- M.- &- Woodward,- James- (1999):- “Independence,- Invariance- and- the- Causal-
Markov-Condition”,-British$Journal$for$the$Philosophy$of$Science$50:-521–583.--
Hawthorn,-Geoffrey-(1991):-Plausible$Worlds:$possibility$and$understanding$in$history$and$the$
social$sciences.-Cambridge:-Cambridge-University-Press.-
Hedström,-Peter-&-Ylikoski,-Petri-(2010):-“Causal-Mechanisms-in-the-Social-Sciences”,-Annual$
Review$of$Sociology$36:-49–67.--
Hempel,-Carl-G.-(1942):-“The-Function-of-General-Laws-in-History”,- The$Journal$of$Philosophy$
39,-2:-35–48.--
---—- (1965):- Aspects$ of$ Scientific$ Explanation$ and$ Other$ Essays$ in$ the$ Philosophy$ of$ Science.-
New-York:-The-Free-Press.-
Hempel,- Carl- G.- &- Oppenheim,- Paul- (1998[1948]):- “Studies- in- the- Logic- of- Explanation”,- in-
Introductory$Readings$in$the$Philosophy$of$Science,-ed.-Klemke,-Hollinger-and-Rudge.-3rd-
ed.-New-York:-Prometheus-Books.-
Hesslow,-Germund-(1983):-“Explaining-differences-and-weighting-causes”,-Theoria-49:-87–111.--
Horgan,-Terence-(1989):-“Mental-Quasation”,-Philosophical$Perspectives$3:-47–76.--
Hume,-David-(1748):-An$Enquiry$Concerning$Human$Understanding,-in-the-series-The$Harvard$
Classics-Vol-XXXXVII,-Part-3.-New-York:-P.F.Collier-&-Son,-1909–14;-Bartleby.com-2001-
- URL-=-<-http://www.bartleby.com/37/3/>-
Jarausch,-Konrad-H.-&-Hardy,-Kenneth-A.-(1991):-Quantative$Methods$for$Historians:$A$guide$to$
research,$data$and$statistics.-Chapel-Hill:-The-University-of-North-Carolina-Press.-
Kaye,- Simon- T.- (2010):- ”Challenging- Certainty:- the- utility- and- history- of- counterfactualism”,-
History$and$Theory$49:-38–57.-
Khong,- Foong- Yean- (1996):- “Confronting- Hitler- and- its- Consequences”,- in- Counterfactual$
Thought$ Experiments$ in$ World$ Politics,$ ed.- Tetlock- &- Belkin.- Princeton:- Princeton-
University-Press.-95–118.-
Kosso,- Peter- (2011):- “Philosophy- of- Historiography”,- in- A$ Companion$ to$ the$ Philosophy$ of$
History$ and$ Historiography,- ed.- Tucker.- Paperback- edition- (hardback- edition- published-
2009).-West-Sussex:-Wiley`Blackwell.-9–25.-$--
Kripke,-Saul-(1972):-Naming$and$Necessity.$Cambridge,-MA:-Harvard-University-Press.-
Kuorikoski,-Jaakko-(2007):-“Explaining-with-equilibria”,-in-Rethinking$Explanantion,-ed.-Persson-
&-Ylikoski.-Dordrecht:-Springer.-149–162.-
---—- (2010):- Society$ by$ Numbers:$ studies$ on$ modelNbased$ explanations$ in$ the$ social$ sciences.-
Ph.-D.-thesis.-Philosophical$Studies$from$the$University$of$Helsinki$31.-
Landes,-David-(1998):-The$Wealth$and$Poverty$of$Nations$–$why$are$some$so$rich$and$some$so$
poor.$New-York:-W.-W.-Norton.-
Lebow,-Richard-Ned-(2007):-”Counterfactual-Thought-Experiments:-a-necessary-teaching-tool”,-
History$Teacher$40,-2:-153–176.-
---—-(2010):-Forbidden$Fruit:$Counterfactuals$and$International$Relations.-Priceton-and-Oxford:-
Priceton-University-Press.-
Lebow,- Richard- Ned- &- Gross- Stein,- Janice- (1996):- “Back- to- the- Past:- counterfactuals- and- the-
Cuban- missile- crisis”,- in- Counterfactual$ Thought$ Experiments$ in$ World$ Politics,$ ed.-
Tetlock-&-Belkin.-Princeton:-Princeton-University-Press.-119–148.-
Lewis,-David-(1973a):-“Causation”,$the$Journal$of$Philosophy$70,-17:-556–567.--
---—-(1973b):-Counterfactuals.-Oxford:-Blackwell-
---—-(1986a):-On$the$Plurality$of$Worlds.-Oxford:-Blackwell.-

106
---—-(1986b):-Philosophical$Papers,$vol.-2.-Oxford:-Oxford-University-Press.-
---—-(2000):-“Causation-as-influence”,-The$Journal$of$Philosophy-97,-4:-182–197.-
Lipton,- Peter- (1990):- “Contrastive- explanations”,- in- Explanation$ and$ its$ Limits,- ed.- Knowles.-
Cambridge:-Cambridge-University-Press.-247–266.--
Little,-David-(2010):-New$Contributions$to$the$Philosophy$of$History,-in-the-Methodos$Series,-ed.-
Courceau-and-Frank.-Dordrecht:-Springer.-
MacDonald,- Graham- &- MacDonald- Cynthia- (2011):- “Explanation- in- Historiography”,- in- A$
Companion$ to$ the$ Philosophy$ of$ History$ and$ Historiography,- ed.- Tucker.- Paperback-
edition-(hardback-edition-published-2009).-West-Sussex:-Wiley`Blackwell.-131–141.-
Mackie,-J.-L-(1974):-The$Cement$of$the$Universe:$a$study$of$causation.-Oxford:-The-Clarendon-
Press.-$--
Malinowski,-Bronislaw-(1935):-Coral$Gardens$and$their$Magic.-New-York:-American-Book-Co.-
Martin,-Raymond-(1992):-The$Past$within$Us:$an$empirical$approach$to$philosophy$of$history.$
Princeton:-Princeton-University-Press.-
Megill,-Allan-(2007):-Historical$Knowledge,$Historical$Error:$a$contemporary$guide$to$practice.-
Chicago:-The-University-of-Chicago-Press.-
Melia,- Joseph- (1998):- “Possible- Worlds”,- in- Routledge$ Encyclopedia$ of$ Philosophy,- ed.- Craig.-
London-Routledge.-URL-=-<http://www.rep.routledge.com/article/N088>-
Mellor,-Hugh-(1995):-The$Facts$of$Causation.-London:-Routledge.--
Menzies,-Peter-(2009):-"Counterfactual-Theories-of-Causation",-in-The$Stanford$Encyclopedia$of$
Philosophy$ (Fall- 2009- Edition),- ed.- Edward- N.- Zalta.- URL- =-
<http://plato.stanford.edu/archives/fall2009/entries/causation`counterfactual/>.-
Miklóssy,-Katalin-(2007):-“Détente-ur-småstaternas-perspektiv.-Fallen-Ungern-och-Rumänien”,-
Nordisk$østNforum,$21,-4:-411–433.--
Mitchell,- Sandra- (2008):- “Exporting- Causal- Knowledge- in- Evolutionary- and- Developmental-
Biology”,-Philosophy$of$Science$75:-697–706.--
Murphey,-Murray-(1973):-Our$Knowledge$of$the$Historical$Past.-Indianapolis:-Bobbs`Merrill.-
Musil,- Robert- (1996):- The$ Man$ without$ Qualities,$ vol- I,- translated- by- Eithne- Wilkins- &- Ernst-
Kaiser.-London:-Vintage.-Original-title-Der$Mann$ohne$Eigenschaften$(1930).-
Nolan,-Daniel-(2011):-“Why-Historians-(and-Everyone-Else)-Should-Care-about-counterfactuals”,-
Philosophical$Studies,$published-online-30-September-2011.-
Northcott,-Robert-(2008):-“Weighted-Explanations-in-History”,-Philosophy$of$the$Social$Sciences-
38,-1:-76–96.--
Oakeshott,- Michael- (1966[1933]):- Experience$ and$ its$ modes.$ London:- Cambridge- University-
Press.$$
Pascal,-Blaise-(1958[1669]):-Pascal’s-Pensées.-Paperback-edition.-New-York:-E.P-Dutton-&-Co.---
Pomeranz,- Kenneth- (2006):- “Without- Coal?- Colonies?- Calculus?”,- in- Unmaking$ the$ West:$
“WhatNif”$scenarios$that$rewrite$world$history,$ed.-Tetlock,-Lebow-&-Parker.-Ann-Arbor:-
The-University-of-Michigan-Press.-241–276.-
Raatikainen,- Panu- (2011):- “Teon- teoria- ja- historiallinen- selittäminen”,- Tiede$ ja$ Edistys$ 36,- 4:-
291–300.--
Raerinne,- Jani:- Generalizations$ and$ Models$ in$ Ecology:- lawlikeness,- invariance,- stability- and-
robustness.-Ph.-D.-thesis.-Philosophical$Studies$from$the$University$of$Helsinki$34.-
Railton,-Peter-(1980):-Explaining$Explanation:$A$Realist$Account$of$Scientific$Explanation.-Ph.D.-
Thesis.-Princeton-University.-

107
Reiss,- Julian- (2009):- “Counterfactuals,- Thought- Experiments,- and- Singular- Causal- Analysis- in-
History”,-Philosophy$of$Science$76:-712–723.-
Roberts,-Clayton-(1996):-The$Logic$of$Historical$Explanation.-Pennsylvania:-Penn-Sate-Press.$$
Ruben,-David-Hillel-(1990):-Explaining$explanation.-London:-Routledge.--
---—-(1994):-“A-Counterfactual-Theory-of-Causal-Explanation”,-Noûs$28:-465–481.--
Salmon,- Wesley- (1971):- “Statistical- Explanation”,- in- Statistical$ Explanation$ and$ Statistical$
Relevance,-ed.-Wesley-Salmon.-Pittsburg:-Pittsburg-University-Press.-29–87.-
Schaffer,-Jonathan-(2005):-“Contrastive-causation”,-Philosophical$Review-114,-3:-327–358.-
Scriven,- Michael- (1959a):- “Explanation- and- Prediction- in- Evolutionary- Theory”,- Science$ 130:-
477–482.--
---—- (1959b):- “Truisms- as- the- Grounds- of- Historical- Explanations”,- in- Theories$ of$ History$ ed.-
Gardiner.-New-York:-Free-Press.-443–475.-
Sintonen,- Matti- (1984):- The$ Pragmatics$ of$ Scientific$ Explanation.- Ph.D.- Thesis.- University- of-
Helsinki.-Acta$Philosophica$Fennica$37.-
Steel,-Daniel-(2008):-Across$the$Boundaries:$extrapolation$in$biology$and$social$science.$Oxford:-
Oxford-University-Press.$$$
Stanford,-Michael-(1998):-An$Introduction$to$the$Philosophy$of$History.-Oxford:-Blackwell.--
Strauss,- Barry- (2006):- “The- Resilient- West- –- Salamis- without- Themistocles,- Classical- Greece-
without- Salamis- &- the- West- without- Classical- Greece”,- in- Unmaking$ the$ West:$ “WhatN
if?”$ scenarios$ that$ rewrite$ world$ history,$ ed.- Tetlock,- Lebow- &- Parker.- Ann- Arbor:- The-
University-of-Michigan-Press.-90–118.-
Taylor,- A.J.P- (1969):- The$ TroubleNMakers:$ dissent$ over$ foreign$ policy,- 1729–1939.- London:-
Panther-History.$-$$
Tetlock,-Philip-E.-&-Belkin,-Aaron-(eds.)-(1996):-Thought$Experiments$in$World$Politics:$locigal,$
methodological$and$psychological$perspectives.$Princeton:-Princeton-University-Press.-
Tetlock,- Philip- E.- &- Belkin,- Aaron- (1996):- “Counterfactual- Thought- Experiments- in- World-
Politics:- logical,- methodological- and- psychological- perspectives”,- in- Counterfactual$
Thought$ Experiments$ in$ World$ Politics,$ ed.- Tetlock- &- Belkin.- Princeton:- Princeton-
University-Press.-3–38.-
Tetlock,- Philip- E.,- Lebow,- Richard- Ned- &- Parker- Geoffrey- (eds.)- (2006):- Unmaking$ the$ west:$
“WhatNif”$scenarios$that$rewrite$world$history.$Ann-Arbor:-University-of-Michigan-Press.--
Tetlock,- Philip- E.- &- Parker,- Geoffrey- (2006):- “Counterfactual- Thought- Experiments”,- in-
Unmaking$the$West:$“WhatNif?”$scenarios$that$rewrite$world$history,$ed.-Tetlock,-Lebow-
&-Parker.-Ann-Arbor:-The-University-of-Michigan-Press.-14–44.-
Thompson,- J.E.S- (1966):- The$ Rise$ and$ Fall$ of$ Maya$ Civilization.- Norman:- The- University- of-
Oklahoma-Press.$$
Tucker,- Aviezer- (1998):- “Unique- Events:- the- underdetermination- of- explanation”,- Erkenntnis$
48:-59–80.-
---—-(2004):-Our$Knowledge$of$the$Past$–$a$philosophy$of$historiography.-Cambridge:Cambridge-
University-Press.--
---—- (2011):- “Introduction”,- in- A$ Companion$ to$ the$ Philosophy$ of$ History$ and$ Historiography,-
ed.- Tucker.- Paperback- edition- (hardback- edition- published- 2009).- West- Sussex:- Wiley`
Blackwell.-1–6.-$-
Turner,-Derek-(2007):-Making$Prehistory$–$historical$science$and$the$scientific$realism$debate,-
Cambridge:-Cambridge-University-Press.-

108
Tuomela,-Raimo-(1980):-“Explaining-Explaining”,-Erkenntnis$15:-211–243.--
Vaidya,- Anand- (2011):- “The- Epistemology- of- Modality”,- The$ Stanford$ Encyclopedia$ of$
Philosophy$(Winter-2011-edition),$ed.-Edward-N.-Zalta.-
URL-=-< http://plato.stanford.edu/archives/win2011/entries/modality`epistemology/>-
van- Bouwel,- Jeroen- &- Weber,- Erik- (2008):- “A- Pragmatist- Defense- of- Non`Relativistic-
Explanatory-Pluralism-in-History-and-Social-Science”,-History$and$Theory$47:-168–182.--
van-Fraassen,-Bas-(1980):-Scientific$Image.$Oxford:-Oxford-University-Press.$$
Vašiček,-Zneněk-(2011):-“Philosophy-of-History”,-in-A$Companion$to$the$Philosophy$of$History$
and$ Historiography,- ed.- Tucker.- Paperback- edition- (hardback- edition- published- 2009).-
West-Sussex:-Wiley`Blackwell.-26–43.-$-
von- Wright- (1971):- Explanation$ and$ Understanding.$ Itchaca- and- London:- Cornell- University-
Press.-
Weber,-Max-(1949):-“Objective-Possibility-and-Adequate-Causation-in-Historical-Explanation”,-
in-The$methodology$of$the$Social$Sciences$by-Max-Weber,-translated-and-edited-by-Shils-
&-Finch.-Glencoe,-Illinois:-The-Free-Press.--
White,-Morton-(1965):-Foundations$of$Historical$Knowledge.-Westport:-Greenwood-Press.--
Winch,- Peter- (1990[1958]):- The$ Idea$ of$ Social$ Science$ and$ Its$ Relation$ to$ Philosophy.- 2nd- ed.$
New-Jersey:-Humanities-Press-International,-
Woodward,-James-(1984):-“A-Theory-of-Singular-Causal-Explanation”,-Erkenntnis$21:-231–262.-
---—-(2003):-Making$Things$Happen.$Oxford-and-New-York:-Oxford-University-Press.-
---—-(2006):-“Sensitive-and-Insensitive-Causation”,-The$Philosophical$Review$115,-1:-1`50.-
Ylikoski,- Petri- (2001):- Understanding$ Interests$ and$ Causal$ Explanation.$ Ph.D.- thesis.-
Department- of- Social- and- Moral- Philosophy,- University- of- Helsinki.- URL- =-
<https://helda.helsinki.fi/bitstream/handle/10138/21811/understa.pdf?sequence=>-
---—-(2007):-“The-idea-of-contrastive-explanandum”,-in-Rethinking$Explanantion,-ed.-Persson-&-
Ylikoski.-Dordrecht:-Springer.-27–42.$$
---—-(2011):-”Social-Mechanisms-and-Explanatory-Relevance”,-in-Analytical$Sociology$and$Social$
Mechanisms,$ed.-Demeulenaere.-Cambridge:-Cambridge-University-Press.-154–172.--
Ylikoski,- Petri- &- Kuorikoski,- Jaakko- (2010):- ”Dissecting- Explanatory- Power”,- Philosophical$
Studies$148:-201–219.-

109